CFS 10x presents
AI for Advice

Good morning everybody.

 

Good morning everybody.

 

I see that the caffeine has not yet kicked in for everybody in the room.

 

Good morning. Also to people who are online as well that I, I can't actually see.

 

I think one thing I'll take away from that introduction is I need to shorten my bio 'cause it'll probably save about five minutes as we go through.

 

So I'm gonna start out with a story, a very different one to the Robin Hood one, I have a 14-year-old son.

 

That's probably not what you'd expect at the start of an AI presentation.

 

My son is highly gifted autistic, which means that he's incredibly intelligent, he's incredibly blunt and he's incredibly obsessive, particularly when it comes to technology.

 

So when chat GPT came out a couple of years ago, it was like Christmas for him.

 

He was using it for absolutely everything.

 

For his schoolwork, for his coding.

 

He was creating images, videos, music, bad music, very bad music.

 

Um, but he was really getting into it.

 

Now, a couple of months ago, I was due to do a presentation at a conference.

 

So I thought I'd check in with him and see how he was getting on.

 

So I asked him, Jake, how are we getting on with chat TBT?

 

And if you have a 14-year-old, you might recognize the meh kind of expression.

 

And I was like, well, you know, are you still using this stuff?

 

And he said, dad, I am. Yes, I'm using it for two things.

 

I'm using it one to get all of the research from my school study so I can really have an impact.

 

And two, I'm using it for complex coding in Minecraft and Roblox to really help me do that because I can't actually do that myself.

 

But what he was really saying to me is, AI is just a tool.

 

That's all it is. It's just a tool.

 

But it really starts to become valuable when it delivers tangible outcomes which are important to you or to your business.

 

And what we're gonna be doing over the next 40 minutes, myself and Jerome, is I'll be talking a little bit about where AI is, where it's going, give you some of the state of the art and Jerome's then really gonna drill down into what does it mean for you and your business.

 

So what does value actually mean?

 

Well, let's take a look at this from Microsoft's perspective.

 

And you might recognize this sound.

 

I think that our industry has to have a common vision.

 

It was a time that connected us to incredible things.

 

My name for this vision is information at your fingertips.

 

And three decades later we find ourselves in a new era.

 

One where access to information becomes access to expertise from the farm to the lab, from the boardroom to the classroom.

 

This new generation of AI is for everyone everywhere.

 

Now anyone can save time with a personal assistant.

 

I'm saving about 50% of time and that's time that I can use to do other innovative things.

 

Anyone can access a personal tutor to learn new skills.

 

I think this technology has a potential to completely reimagine the way every single student learns in the world.

 

This is a new way to analyze with a personal coach.

 

We're gonna be able to have not only productivity gains, but insights served to us near real time.

 

Generative AI can learn from the data to help improve the farmer productivity.

 

AI is unlocking creativity for us. All descriptions are so detailed in my imagination. I can paint the artwork with expertise at your fingertips.

 

You can build what matters.

 

Is there anything else you need help with?

 

Welcome to the age of AI transformation.

 

So what you saw in that video throughout were different uses of the same generative AI technology.

 

But as I said at the start, it's all around the value which is applicable to you in your business.

 

So how do we think about this from a business perspective?

 

So at Microsoft, when we work with our customers, no matter what type of technology it is, and AI is definitely a great example of this, we normally look at value in four areas.

 

The first one is how it can be leveraged to improve your employees and their experience.

 

The second is how you can engage better with your customers.

 

The third area is really thinking about how we can reshape, automate, and define business processes to optimize them.

 

And there was some examples right at the very start in the presentation.

 

And the fourth area is really looking at how do we attract new business?

 

How do we get new revenue?

 

How do we look at innovation?

 

So as an example in the insurance industry, we could use generative AI to fully automate the insurance claims process, look at fraud detection and have no human in the loop at all, which is completely different to what we might see today.

 

There are a couple of other areas as well to really think about for sustaina, uh, for uh, ai.

 

One of those is sustainability.

 

So we do often hear ai when you roll it out, it's gonna take all the energy and all the water and blah, blah, blah.

 

But the two sides to this coin.

 

One, we need to mitigate alienates potential impact, but we can use AI for sustainability use cases for next generation power utility grids for modeling things like climate change for sustainability reporting.

 

But equally on this side also accessibility.

 

How do we think about your workforce and making sure we're using AI to ensure that your employees are getting the right experience or your customers are getting the right experience.

 

I mentioned my son earlier, he has a challenge with things like social cues.

 

So we can actually run AI across our teams calls and in real time it can actually monitor that call.

 

It can actually pick up the sentiment of the different parties in the conversation and can feed to a neuro diverse person what's actually happening in the conversation and even give them the type of questions that they might want to ask to join in as part of that.

 

So think about that for inclusivity in your workforce or for your customers.

 

Now drilling this down a layer deeper and drone's gonna go into some very specific examples in a lot more detail for you.

 

But where we've worked with financial organizations all around the world, we typically see four key areas where we're seeing generative AI success today.

 

The first one is around that employee experience.

 

So looking at things like employee wellbeing, we can use AI to do things like predict are they working strange times, are there lots of inaccuracies in the work?

 

Which might mean there's a personal problem looking at things like, uh, employee enablement.

 

How can we make sure we're doing real time skilling?

 

How can we do things like on the job coaching in real time using AI and looking at things like employee productivity.

 

So really getting rid of those menial tasks that you might have to do around note taking email generation and so on.

 

So you can work on things which are more customer facing or more valuable to your organization.

 

So as an example, AI life insurance and what they've really started to do, thank you for the loud noises there with the ice cubes.

 

Uh, so a ai uh, life insurance, they're really focused on automating a number of processes in the business, like email generation, content generation, automatically for their clients.

 

And that's really freed up their customer facing agents so they can focus on activities which are customer facing and revenue generating.

 

The Bank of Queensland as an example, uh, they're doing very similar things, but they're seeing between two and a half and five hours of freed up time per week just by automating some of these processes.

 

The second area is using some of the data you can get from your forms, from your processes inside of the business to get insights.

 

So looking at things like how can we do better project management, how can we look at contracts, legal and risk management using AI to help with that process and looking at things like document analysis.

 

So the national bank in Greece, they're actually ingesting all of the forms, all of the paper forms they get across their entire business as they come in.

 

And they're able to automatically use generative AI to classify the forms, get insights in terms of the information that's on those forms and feed that data real time to their customer facing agents in real time.

 

And what this is actually doing is giving them a 90% accuracy in terms of the speed and the accuracy of the data that's given to their frontline people to deal with clients.

 

The third area is around client engagement and sales.

 

I think everybody wants to have happy clients and more sales.

 

Um, but there's types of things you can think about.

 

Personalized customer relationship management, looking at opportunities for upselling cross-sell when you're in a conversation, having an a agent or an AI prompt you to do that.

 

And also looking at things like knowledge management, ally Financial, um, one of the financial agencies in the uk, they are typically doing two things when they're working with their, with their customers or their clients.

 

The first one is actually talking to them about what they need to do, what type of investments they can do and so on.

 

The second then is actually taking all of the notes and writing it up after the call to take actions.

 

So what they've decided to do is run, uh, uh, AI across their teams calls.

 

When they engage, it means they can actually be more engaged with the person on the call.

 

The whole of the transcript is captured.

 

All of the call notes are summarized, the actual actions are taken and prompts to take better action or produced in real time.

 

And what this is actually doing is giving them up to 50% more time due to not having to do post-call notes to actually deal to provide better customer service experiences.

 

And the first final area really around things like content generation and management and marketing.

 

So Bank of Queensland, again, they're actually creating, um, real time marketing, uh, content using generative ai.

 

They're creating, uh, HR and and legal, um, uh, legal documents in the background.

 

They're also doing things like job descriptions and so on, all automated and generated by AI speeding up the processes.

 

So you heard about a lot of things that potentially we can do, but why is generative AI so exciting at the moment?

 

Why is everybody talking about this as being the era of ai?

 

AI is not new.

 

It's been around since the 1950s.

 

We had machine learning in the 1990s, deep learning in the 2010s.

 

But generative AI came out in the 2020 early 2020s.

 

And I'd say now we're the early stages of generative ai and we're still on this journey towards what people call general ai when you can't really distinguish software and machines versus a human.

 

There's no terminators coming yet though, I just point that out.

 

But what we have seen is, for me personally, the fastest acceleration of any technology that I've ever seen in my 25 plus years in this space.

 

Now, there are three main things for me that generative AI givers that no other technology has done before the first, it can actually adapt to us as humans and learn about as in real time and give us personalized experiences.

 

Two, for the first time it can create net new content.

 

I talked about my son, but it can create net new videos, images, text, it can create a marketing campaign for you if you wanted it to do that.

 

And three, for the first time, we're actually able to talk to AI using human language.

 

I can actually use human language to take advantage of all of these data sets across an organization.

 

I don't need to be a complex coder to take advantage of it.

 

And what we've really seen over the last couple of years is this acceleration in the capabilities of AI itself.

 

So we've gone from having text conversations where someone responds back to us, we have chat bots that can interact with us in real time and can actually transform the conversation and we can interrupt them, it can speak back to us in real time.

 

It can adapt to us.

 

We've had code generation that I mentioned, but really most of the focus today and where we're going in the future are on these two areas highlighted in yellow.

 

You may have heard of agent AI.

 

Has anyone heard of agent AI?

 

That means raise your hand if you have.

 

So a few people in the room.

 

Okay.

 

So agent AI is really the concept of having software agents that can automate business processes.

 

So you might want to onboard an individual today.

 

You might say, oh, we need to order a laptop, we need to get this piece of training, we need to get access to this system.

 

An agent can just do that using software, freeing your time up to do something which is more impactful.

 

So we're seeing a lot of focus on agent AI.

 

But the second area is really on this right-hand side, which is around the human element of AI.

 

So the models themselves have really advanced and what we're seeing now is a lot more capability around things like reasoning and problem solving, sentiment, really understanding human language, human experience as you engage with it.

 

And the third one is inference.

 

Being able to extract relevance from something and making it valuable.

 

The fact that we can also interact with AI using text, we can talk to it, we can use video, we can show it images.

 

When you start to wrap all these things together, we're starting to have much more of a human experience with AI.

 

And as people understand human experiences, they also then start to understand how can we leverage AI to help augment human capabilities inside of an organization?

 

And this is really why it's taking off.

 

Now I'm gonna show you a video or a demo of the art of the possible today.

 

This is really looking at a real-time customer service agent interacting in different ways with a human being.

 

Um, I'm not gonna expect you to go away and build one of these.

 

You can have a bit of a play with no code and low code later, but it's to give you an idea of the art of the possible.

 

But as you go through this halfway through, just take a look to see how the agent itself is built.

 

We do this using human language.

 

We don't have to code it.

 

I wanna make sure that I am ready and geared up for this camping adventure with my friends.

 

So I've started a chat here with Kentoso Outdoors.

 

Notice this is something that is now routine AI interacting with me in the chat and with images.

 

But now I'm gonna ask it to do something a little bit more because we want these AI agents to break out of the chat box and bring real-world experience to all of us maybe with something conge like this.

 

Alright, let's talk to it.

 

Hello, are you there?

 

I said from our chat, I see you're getting ready for a winter camping trip and need help with cold weather gear.

 

You shared a snowy forest image, so it sounds like you're preparing for some serious cold to assist you further.

 

Can I send the concierge page to your browser to give you more detailed advice?

 

I think that would be delightful.

 

Let's do it.

 

Great.

 

Sending the concierge page to your browser.

 

Now this will help us find the perfect cold weather gear for your winter adventure.

 

Let's get you all set up for a cozy and safe trip.

 

Awesome.

 

Well, what do you recommend, uh, for me to buy?

 

Great question, Seth.

 

Since you've already got the Skyview two-person tent, let's ensure you're fully prepared.

 

How about the mountain dream sleeping bag?

 

It's perfect for cold weather with a cozy design and temperature rating down.

 

So this is all great.

 

Is there any way you can write this up in a way that I can actually take a look at it?

 

Absolutely, Seth.

 

I'll put together a personalized recommendation article with all these items.

 

Once it's ready, it should pop up on your screen for you to review.

 

Give me just a moment to prepare that for you.

 

That sounds great.

 

Lemme know when you're ready.

 

Now here's the thing.

 

The way that you actually code this is with English because these are the prompts that we're giving the AI agent to help us to help it understand how it needs to react.

 

And again, this is all real code.

 

You can see I'm literally running, Seth, I'm just finishing up your article now.

 

Alright, you should see the personalized recommendation article on your screen.

 

Take your time to check it out.

 

This is awesome.

 

Thank you so much.

 

We'll talk soon.

 

Hopefully.

 

You are welcome, Seth.

 

Enjoy your winter adventure and feel free to reach out anytime.

 

Talk soon.

 

How cool is that?

 

And remember, this isn't a peek into the future.

 

It's happening right now with Azure AI Foundry.

 

Back to you, Satya.

 

Now, I'm just gonna say we're not handing over to Satya, who's Microsoft's CEO.

 

That would've been a nice dial-in today.

 

But I really wanted to give you an idea of the capabilities of the technology itself today.

 

What you saw there was voice interacting with text, interacting with web, pulling back, real-time search.

 

You could interact and interrupt the conversation itself.

 

So the technology's really advancing.

 

But again, I'm not gonna expect you to build this when you have your play later on today.

 

I also wanna keep the conversation very grounded and Jerome and I were talking just before the session about AI itself because what I typically see in many organizations that I work with is this concept of the proof of concept graveyard.

 

So they have a bit of a play with the technology and you have a science experiment here, a bit of a play there, but there's no real strategy to actually leverage the technology.

 

So it fails in these little small pockets of proof of concepts.

 

And what we're really starting to see because the technology matures so quickly, is people revisiting.

 

So I really advise you to give it a go.

 

But what we have really noticed is that organizations that move beyond that proof of concept graveyard and embrace this at scale inside of their organization, follow four key areas.

 

One is really looking at how we have the right leadership setting that strategy to say, we are gonna embrace this.

 

We are gonna use this.

 

It's an opportunity for us to leverage in our business.

 

The second one is really around the human change.

 

As we adapt and we have AI doing things for us, we have processes automated and so on, we interact with our customers in different ways.

 

It also means that we need to have the right skills, the right knowledge for the employees.

 

It also means we need to have the right type of change agent to make sure this happens successfully.

 

The third area is around having the right technical readiness.

 

So the right types of data sets inside of the organization, the right infrastructure, but more importantly, the right people who can understand what does AI actually mean for you, your use cases and where it generates value.

 

And then the bit down at the bottom, we always need to consider things like security, responsible and ethical AI and governance.

 

And the governance is probably the most important thing because as you deploy AI in your business, you wanna make sure you have the right guardrails, the right capabilities, procedures and processes in place to make sure it's doing what it should be doing.

 

Or you can remediate against that.

 

Now I've just got a couple more slides and we'll, I'll be wrapping up and handing over to Jerome to go deeper into this for you.

 

We ourselves at Microsoft have rolled out AI, traditional AI and obviously the last couple of years, generative AI across our entire business.

 

And you can start to see the types of statistics and information that we're getting fully across the space.

 

There's a link and all of the slides today will be provided, but there's a link at the bottom, which actually is a, a study that we've done where we've deployed this across our business, but we actually share the learnings with people to say what worked?

 

What value did we get from it?

 

And more importantly, what didn't work?

 

What were the challenges?

 

How did we overcome those in an organization?

 

So I'd highly encourage you to read that, but to give you an example, one of the areas we've worked on is service, customer service.

 

So now when organizations call into US businesses, around 30% of all of those calls that come in are automated fully, no human in the loop.

 

75% of consumers like us that might have an Xbox problem that calls in, they're handled 75% fully autonomously.

 

And what this really means is two things.

 

One, there's a much shorter resolution time for people dialing in.

 

They get their answers quicker.

 

And two, it actually frees our service agents up to focus on those media problems and provide a better customer service experience.

 

So I'm gonna leave you with a little bit of an art of the possible and I'll come back for one last statement.

 

Have a look at this one.

 

The king in residence at the moment.

 

Yes, it looks like the king is in residence right now.

 

The royal standard flag is flying above Buckingham Palace, which is the signal that the monarch is present.

 

How exciting to be there while the king is home.

 

Try and tell me exactly what they're doing right now, please.

 

Um, right now the ducks are gently gliding across the water.

 

They're moving in a fairly relaxed manner, not in a hurry.

 

Occasionally one of them will dip its head under the water, probably looking for food and then pop back up.

 

I even know when a taxi is coming with its orange light on.

 

I think I'll hail it to get home.

 

Yes, I spotted one just now.

 

It's heading your way on the left side of the road.

 

Get ready to wave it down.

 

Okay, great job hailing that taxi.

 

It looks like you're all set to go.

 

That's a good dog right there leading the way into the taxi.

 

Safe travels.

 

So just an idea of the capabilities again, and we don't always have to think about these in the use cases that you see them, but the same capabilities could be used for training, inference, extracting information.

 

I would probably with that one, they'll have a less patronizing voice at the end.

 

Congratulations.

 

Getting it.

 

Taxi is not that exciting.

 

So I'm gonna leave it with this last thought.

 

And this is really important.

 

Generative AI is still only at its early stages of development.

 

And you've seen the capabilities we already have today.

 

The technology itself is only gonna improve, it's gonna get smarter and it's gonna be infused more and more in your daily lives and in your working lives.

 

So the key thing for me here is that you do not want to get left behind.

 

A lot of people say, Rick, will AI be replacing humans?

 

I would say definitely not in the short term.

 

I'd actually say the threat is much more likely to you to come from another human that's using AI to get business advantage when you are not using it.

 

And the second thing from a business, you're much more likely to suffer from competitors who are using AI or new startups who are using AI for competitive advantage.

 

So please embrace it.

 

I'm now gonna hand over to Jerome, I think, or you come back up first.

 

I'm gonna hand back over here.

 

But thank you very much for listening and enjoy the rest of the day.

 

Thank you.

 

Um, I'm just going to get straight into this.

 

So, um, we're just running a couple of minutes behind.

 

Um, so Roan, um, please, uh, join us on stage.

 

Roan is our, uh, CTO in our business.

 

Um, and prior to joining CFS Jaran was a COO at East Spring Investments in Singapore, partner in EY in Hong Kong, led the APAC wealth and asset management business.

 

So, um, he's been everywhere around this world.

 

Um, and we also have a very special guest, Taylor, uh, Phillips, who's, uh, joining us from Newcastle from Hunter Financial.

 

And she's gonna be speaking about her experience, uh, using AI in her business.

 

So please join me in working them both to this stage.

 

Hi everyone.

 

Obviously Michelle, that resume was generated with Jen AI.

 

Yeah.

 

So, uh, um, maybe, um, show of hands to start with.

 

Who, who's used Gen AI Hanza.

 

Good, good.

 

So we're talking to an audience that's got a little bit of experience.

 

Now.

 

Rick already mentioned some of the capabilities and, and interesting you mentioned your son.

 

I think we can all relate to that, Rick, because, uh, we all have got kids that have been using chat GPT for their homework and for other things.

 

And in my case, uh, Quentin, my son, he, um, he, he's been passionate about GPT using it for everything.

 

So last year I explained to him that, uh, I was aiming to use Gene AI, uh, to support Craig Day and his technical services team.

 

Now you all know Craig Day, and thank you for those voting for him, as Bryce said to, uh, to be the leading technical services team for 12 years running.

 

So well done Craig on that one.

 

Now, he said to me, Quentin, he said, dad, if Che GPT can pass the bar exam and medical exams in the US, surely it should be able to answer some questions around the Australian tax and superannuation legislation here in Australia.

 

So logically, I put 'em straight to work with Microsoft and Avanade I thought was a good way for him to earn back his international education.

 

And, uh, what we were trying to achieve, we were trying to scale the team.

 

Uh, Craig and his team are managing in a very complex environment, insatiable amount for his services and a very long lead time to get experts up to speed.

 

And so, so we were aiming to see how, how Gen AI could help answer questions for, for Craig and Esteem.

 

Now, how did, how did we go on that?

 

Uh, we, we coined the first tech bot and, uh, and there's no better, better person to talk about it than Craig.

 

So let's hear whether the bot was able to beat Craig.

 

Craig, why don't you come on the site?

 

Thanks everyone.

 

So for my team, we deal with an awful lot of complexity.

 

As you as advisors, we get about 12 to 15,000 calls a year on everything from SMSF to wage care, to tax to superannuation.

 

And my guys need to be skilled up to be able to understand the question and respond to it and give you some ideas about how the rules work in that situation and what to watch out for.

 

As Jerome said, it takes me about 18 months to two years for a new person really to be useful for me during that time.

 

We are training, training, training, training.

 

So we thought this bot, let's look at this bot and say, how can we use that to assist one of our guys to skill up to do this job maybe a bit quicker, but also in the longer term, how could we potentially think about pushing that out to advisors?

 

So you go on and you get through the first tech chat bot, and basically what we did is we loaded in all of our own content.

 

So we've got about three to 5,000 pages worth of technical gibberish.

 

Basically very, very technical staff strategy, how the rules work, transfer balance, capital aged care stuff, right?

 

Um, and give you the ability to actually ask a question and for it to produce a response.

 

So where do we get to?

 

So we've got an example of the first tech bot, uh, hopefully in the next slide.

 

Yep.

 

Um, and so what you could do in there is go in, ask it a question and what did we find?

 

Initially it wasn't very good, right?

 

But what I describe as both amazing and terrifying is it didn't take very long.

 

It was a matter of weeks of pulling different levers and organizing the data in a different way before we were getting it to about 65% accuracy.

 

And I'll give you an example of one of the common types of questions.

 

Don't try and read this, right, the answer, the question is, my client recently sold their home.

 

How will the sale proceeds be assessed, assessed for social security purposes?

 

A very common question that we get in the team and after a couple of weeks it can produce a response that if you read that, you go, geez, that's pretty good, but it's, when you sit down and look at the detail, you go, Ooh, there's some stuff missing in there.

 

So I would actually rate that a 6.5 out of 10, that response.

 

And it's not because it's actually coming up, well, there's one major error in it, but there's, it's mostly what it's missing.

 

So if my guys gave that response, there's nothing wrong here, but if you as an advisor relied on it, you would go down a channel where you're missing potential opportunities or putting the client in a position where you wouldn't get the outcome that they wanted.

 

So can we get it to the point where it's really reliable?

 

Well, I'm sure as we just saw where in the very early stages, but what do we need to do to get it to that stage?

 

We actually need to get it to learn to listen.

 

And once we started to think about, well, how does it listen?

 

It listens to our calls, starts to transcribe them, that actually opened up a whole bunch of different thinking for the bot in terms of this thing is a tool, it's not gonna replace us.

 

It allows us to do things more efficiently, not only for us, but also for advisors.

 

So we started to think about, well, if we're actually transcribing the call, so it can look at the question and the answer and start to learn from that, what else does that do?

 

Well, we could use potentially agent AI where we actually get off the phone, it produces a summary for us.

 

We check it, press a button, and you've got a summary of your conversation with us in your inbox within five to 10 minutes of getting off the phone that goes straight into your file note.

 

What does that do?

 

It gives you a really good summary of your conversation with us, but it's also addressing issues around risk and compliance because you've now got a trail that shows the type of questions that you ask and the issues that came out of that.

 

Now, not only do we start to think, yeah, that'd be good, but one day someone of my guys said, well, you know, how many times does an advisor go to a CPD session and they're sitting up the back listening to Craig bored out of his brain talking about something you're not interested in, and you get your CPD points.

 

Well, in this situation, you've now sat with one of my guys on the phone for 20 minutes going through a complex issue.

 

We've explained the rules to you, we provide you a summary.

 

Could we have a chat bot that looks at that discussion?

 

Says, here's your CPD points out of this.

 

Also, we get the bot reference library so it's beginning to learn.

 

So it's taking all of our content and looking at common questions and the answer that comes out so it doesn't miss the things that I talked to you about before.

 

Not only that, it can look at your potential business and say, okay, I've got 10 advisors where your strengths and where are your opportunities in terms of your technical knowledge and feed that back into your business.

 

And not only that, finally for me, it provides risk management because once this bot gets good, I can have it listening to every single one of my calls and it can spot the one that doesn't sound right.

 

And we can go back in and have a look at that currently for quality control.

 

Yes, you would be glad to know that we do do quality control on the first tech calls, but we can do a tiny minority of the number of calls.

 

All of a sudden we get a hundred percent of the calls being checked and then validated when something pops up.

 

So we're checking by exception, not looking for a needle in the haystack.

 

Now, for you as advisors, I know it's a common thing.

 

What you do is you call First Tech and then you call another tech team and another tech team and another tech team, and they all same say about the same thing.

 

So therefore I'm pretty confident in the answers I'm getting.

 

What if you just needed to do it once?

 

So that's where we've got to in terms of, uh, first check bot development.

 

Hopefully we'll continue to work on this over time and we can actually begin to develop and deliver really good outcomes for you guys where you can use it as a tool just like we see the ability to use it for the first Check team.

 

Oh, thanks.

 

Thanks, Greg.

 

It's been an absolute, uh, absolute pleasure to work with you and your team.

 

And I think we're now up to about what 60, 60, 60 5% of questions that we can answer through the first deck bot.

 

So, uh, more to come.

 

Let's have a look at what's behind that.

 

So Stanford in the US University did a study last year where I reached two conclusions.

 

One, the advancement of AI capabilities is experiencing rapid acceleration.

 

Acceleration.

 

And two, AI capabilities have reached human parity where they can match and beat the, the average human on many dimensions already.

 

Now, if you're ever wondering what all the hype and speculation is about, it's that speed at which AI is developing and the prediction as to what that's gonna mean if you extrapolate those lines in the next few years.

 

However, as per the first tech bot Advancement is being achieved.

 

But getting to and beyond experts like Craig and his team is hard.

 

We've experienced it and many other organizations are experiencing that as well.

 

Those last percentage improvements, uh, require a lot of work.

 

Now, on the other side of the, uh, of the US, uh, at MIT, a specialized team under leadership of Professor Andre Low has been able to get close but unsuccessful in leveraging generative AI in order to provide trusted financial advice.

 

To date, Professor Low expects that his ongoing research will show, uh, that complementary modules or finance-specific large language models will be continue to be required, uh, to navigate the complex legal, ethical and regulatory landscape that we operate in.

 

Just reflect on that, a specialized, well-funded team's been going at this for a few years and they haven't been able to get there.

 

So they believe it can be done, but they haven't been able to get there yet.

 

One of the big hurdles, I believe, is that, um, generative AI is inherently probabilistic.

 

It is very good in guessing the answer.

 

Whilst if you wanna do text or investment optimization, you probably require something like stochastic modeling, where you use mathematical models like Monte Carlo analysis to approach the answer.

 

And regulators, as you all know, often require, uh, a deterministic approach where the answer is logically derived.

 

So there's a mismatch in terms of paradigm between AI and what regulators expect of you.

 

Furthermore, BBC published an article about two weeks ago, uh, where they had AI summarized news articles that, um, copilot Jet GPT, Gemini, and I think perplexity, uh, summarize a hundred articles.

 

And then I had a journalist with expertise in the field rate the quality of the AI answers, uh, that came out.

 

51% of the answers had, uh, significant issues, accuracy, uh, position 19% of the answers introduced factual errors.

 

So if you, like some of my staff members, use Jet GBT to write a birthday card for your partner, yeah, make sure to check it.

 

You don't wanna get the aids wrong, yeah, on, on those steps of things.

 

So, uh, so that's a kind of seemingly contradictory views, all credible sources.

 

So where does it leave us as Satya Nadela?

 

I think, um, uh, Rick said, um, humans in an AI environment are either in the loop where we participate in a task, on the loop where we monitor and check a task, or off the loop where AI is doing things autonomously.

 

In financial advice, AI capabilities will augment the advisor well before they will be successfully, directly employed with, with customers.

 

And so this will be an evolution rather than a revolution.

 

Now, irrespective of all that, a huge amount of money is flowing into our industry, both globally and locally.

 

I see some very big names there.

 

Investment managers obviously have been using AI for a long period of time already.

 

They use deep learning and machine learning to underpin their, yeah, low latency, uh, algorithmic trading.

 

Uh, I thought I might touch today on two examples.

 

The first one is Morgan Stanley.

 

They have a partnership with OpenAI, uh, the global leader in AI that, uh, the inventor of chat GPT and also sit behind copilot.

 

So let's have a look what Morgan Stanley and OpenAI are up to.

 

How do you create more capacity?

 

How do you allow our financial advisors to have better conversations and more of them?

 

We're a global organization.

 

We have 80,000 plus employees, we have a lot of information available to employees, but their ability to be able to effectively find that information was always a challenge.

 

We were finding that employees were only able to find about 20% of the documents in our whole knowledge repository.

 

We recognized that maybe this was a way to take our knowledge base that we have and give it to the AI and see if that could almost democratize the access to the information.

 

And when we hooked up the artificial intelligence into our assistant, we actually found that now the coverage is about 80% of the documents have been clicked on.

 

And then the question became, where do we point, you know, where do we go next?

 

So after we had success with the assistant, the first thing that came to everyone's mind was a meeting summarization tool.

 

And especially when you think of our environment at a place like Morgan Stanley, when you are meeting with clients, those meeting notes are critical.

 

So the feedback we're getting from clients is really positive.

 

Now they're getting a follow-up email the same day, sometimes an hour after the conversation.

 

And so they feel like the customer service level has increased that their advisor is more engaged with the conversation that they're having and that they're able to kind of have those actionable key takeaways immediately.

 

So Morgan Stanley and OpenAI.

 

Now, it's not hard to imagine how one of their 15,000 advisors uses Gen AI to look through their historic client interactions.

 

Looks at the investment portfolio, uh, uh, performance of the, uh, the investment portfolio sources, news and home, uh, home, uh, uh, views on, on that investment portfolio, bundles that all together in preparation for their next client meeting or even used an AI during the client meeting in order to answer questions that the client may have an interest in.

 

And then subsequently use it after the meeting to summarize as you saw in the video, the meeting, and get the file notes back, uh, into the repository and the summary back to the client.

 

Now this, you don't need to be a Morgan Stanley with 80,000 employees to do these types of things.

 

Microsoft platforms have capabilities today that you can use to do a lot of these types of things.

 

And later in the workshops you will see very practically, uh, Dan and others will show you how to use these types of tools to start introducing some of these capabilities in your practices today.

 

Now, um, uh, the next example I think that we might go through is, uh, origin.

 

So the last one, this is a startup, a startup in the US.

 

Um, and as you can see, origin here is, uh, an easy to use customer app.

 

Now, Australian banks already offer spend analysis on their online portals and mobile apps.

 

Today, what origin does it combines spend analysis, but budgeting, adding targets to your, to your spend and investments in, in a very easy app.

 

As you can see here, they use generative AI to dynamically personalize, engage and nudge with their customers.

 

As you can see here on the picture, they show the investment portfolio dynamically the performance on the investment portfolio, and then they source news relevant to show what the top movers are and why those top movers potentially have moved.

 

Now, when you have got good eyes, you can see down the bottom a button that says, chat with my planner.

 

Because despite all of this fancy technology, people are still interested to talk to a human expert.

 

Yeah, they still wanna see the proverbial widening the eyes if only to confirm some of the assumptions that they have.

 

Now, we've been talking to a lot of planners like you in Australia over the last kind of year or so, and I know that a lot of you have used Gen AI and have been really disappointed with it because for the first time when you use it, you don't quite get out of it that you, uh, that you, uh, that you want to.

 

And typically you only have one chance to make a good impression.

 

But in this case, I urge you to try again because the models will become better every day, every week, every month.

 

And more importantly, you will become better in asking the right questions.

 

And again, we'll go into prompting later on to show kind of the way in which you can ask questions to get the most value out of, out of generative AI.

 

But rather than me talking about what we've learned from all the planners, we've got Taylor Phillips here, uh, from Hunter Financial.

 

Please come on to the stage round applause for Taylor.

 

Hey Taylor, welcome.

 

Thank you.

 

Thanks for joining us today.

 

Really appreciate it and, and really nice to spend a day in, uh, in Newcastle with, with your team.

 

So thanks for, for having us over.

 

Great partnership.

 

Um, maybe start Taylor with, um, giving a bit of an overview of Hunter Financial and your role.

 

Yeah, of course.

 

So we are based in Newcastle, New South Wales, just a few hours off the road.

 

And, uh, the company was founded in 2004 by Philip Smith.

 

Some of you may know Phil.

 

Uh, we're licensed by Account Financial and we are currently a team of 28, which includes four advisors.

 

We're holistic goals-based advisors and our success in that space would place us in the top 10% of firms in Australia.

 

Uh, I've been with the company for 10 years and long done, yeah, long done.

 

And, uh, in that time, the, uh, client numbers have more than doubled.

 

We're servicing about 780 clients at the moment across those four advisors.

 

And our revenue's done the same and some more.

 

So we've come from a top line revenue of 2.8 million, uh, 2015 to 5.7 in 2024, and we're on track this financial year to 7 million.

 

So we're pretty excited about that.

 

I lead the business support team, which oversees the operations and the resources in the business, and that naturally includes the tech and, uh, systems.

 

Lucky you, huh?

 

Yeah.

 

Yeah.

 

Good.

 

So much fun.

 

Um, look, look, thanks and congratulations on what has been a remarkable journey over the last couple of years, obviously.

 

Uh, maybe, um, touch on some of the challenges that you faced, business challenges that led you to using GenAI.

 

Yeah, so, uh, we didn't go looking for AI as a solution initially.

 

Uh, it really found us and, uh, that came from the business having big goals.

 

Um, we know that there's a growing need for advice in Australia, and we wanna continue serving our wonderful clients whilst engaging many more new clients.

 

We wanna increase our capacity without necessarily increasing our, uh, human resources.

 

'Cause as you can see or have seen, we've got a large team.

 

The support ratio is quite high, but we believe the future of the industry is really bright and, uh, that is going to be including a lot of technology.

 

So our business planning kicked off a tech review.

 

Initially, we were just focusing on very common challenges like improving our cybersecurity, uh, upgrading our devices for the next generation of technology and, uh, ultimately moving from servers into the cloud.

 

Yeah, so once we got those foundations, uh, we then, uh, had AI presenting itself as opportunities for really quick and easy wins to automate our repeatable administration tasks.

 

Excellent.

 

And, uh, if you don't mind, I'm gonna use that.

 

AI found us.

 

You weren't looking for AI, AI found us.

 

Yeah, that's what you said.

 

Good.

 

I'm gonna say my next presentation that's, uh, uh, look, digging a little bit deeper on that.

 

So you implemented Microsoft end to end, um, started to find lots of opportunities around AI.

 

Maybe talk through some of the, and I think on the screen we've got some of the capabilities that you have, uh, that you have created.

 

Yeah, absolutely.

 

So in the last six months in particular, we've implemented five key things that have really, uh, laid the foundations again for, uh, the more advanced AI solutions that we're now looking to implement.

 

So we started with SharePoint and OneDrive, that's the Microsoft cloud-based document storage solution.

 

It's, uh, important to spend some time in this space to get your security perimeters right and prepare for the use of Copilot.

 

We didn't want our staff to be able to access private and sensitive information in HR files in particular.

 

So this has allowed more collaboration and communication across our team.

 

It's also reducing the risk of data loss or people saving over your work, which is no longer an issue in our office.

 

Good to hear.

 

Yeah.

 

Second was Teams.

 

Uh, we were late to the party joining the Teams, uh, when COVID really kicked off.

 

Uh, we are using Teams now in everything that we do.

 

Phone call, transcripts, meeting transcripts and recordings even for the appointments that are face to face.

 

That helps with the, uh, face-to-face contact with the, uh, clients or the online meetings during the appointment and post appointment.

 

Uh, especially on that.

 

Um, are clients comfortable?

 

Yes, of course.

 

You would always ask for consent.

 

That's a given.

 

Uh, we haven't had any kickback there.

 

Everyone is very comfortable.

 

So, and we've received really good feedback from our clients about the, you know, the efficiencies that they are seeing post appointment.

 

So that really reinforces that we're doing the right things.

 

Excellent.

 

Yeah.

 

Uh, so look, big journey.

 

Mm.

 

Lots of implementation.

 

Um, if you were to say, what were some of the, the challenges or the benefits coming out of all of this?

 

Yeah, so we also, before we jump into that, if you don't mind, go for it.

 

No, no, no, you're wrong.

 

You jump the gun.

 

Uh, so we really spent a lot of time in Copilot.

 

Oh yeah.

 

Uh, Copilot for us is helping our advisors and our management staff in appointments.

 

Uh, generating file notes is one thing.

 

Analyzing tone is great, especially for HR related meetings, is providing advice summaries, it's analyzing or comparing documents.

 

It's helping us improve our cybersecurity in this space too.

 

Uh, so we have control over the AI that our staff are accessing, which is great.

 

Uh, we also use Forms and Power Automate, which is, we're making our way down the, down the slide, uh, to manage our annual client event.

 

So in November we hosted our annual client event, which was a very similar room to this actually about 120 clients attended that event.

 

And Forms and Power Automate, manage the, uh, registrations, the thank you emails, personalized content post meeting, and a request for feedback.

 

It's now gathered in a way that we can actually use, it's gonna help us plan the next event, which is amazing.

 

'Cause we are advisors, we are not event managers.

 

Uh, and then last but possibly most exciting is Bookings.

 

So we, uh, um, rolled Bookings out in January.

 

That's the Microsoft online scheduling tool for staff and customers to book appointments with ease.

 

So we don't have this back and forth of, uh, of emails that happen to book an appointment.

 

Hey, I remember that when we were in, uh, Newcastle, that we, we looked at the booking system, so you've gone live with it.

 

So we have a quick look at how that looks like.

 

Yeah.

 

All right.

 

So this is the end result.

 

We, the setup was quite, uh, easy once we played around with some settings.

 

When your team was in our office, there was a few quirks that we had to iron out.

 

We used Power Automate to do a lot of that ironing.

 

Uh, and we have rolled it out, uh, with still a bit of a wishlist of things that we would like to achieve, but we are really aiming for progress and not perfection.

 

We have an average of 140 clients per month.

 

So even if we can win a, a minute here or a minute there, it's adding up to a lot of time for our team to spend elsewhere.

 

Yeah, of course.

 

Yeah.

 

So challenges and impacts.

 

Yeah.

 

There we go.

 

Do you wanna cover that?

 

Yeah, yeah, please.

 

So, uh, some of the challenges, uh, or the biggest challenge that we've faced is the people.

 

It's not the tech, it's the change management of course.

 

Um, so the rolling out of the new Pro process, the adoption of it and the training attached to it has been, uh, significant.

 

But I think we're over that hurdle now and we're into a new phase, which is very exciting.

 

For me in particular, the, the impact has been, um, vast.

 

I'll try to summarize it.

 

So firstly, with teams and copilot, we have increased a cap.

 

Uh, increased our capacity for client appointments by that 23%.

 

So for us, that's about 270 appointments that we've completed this financial year, year to date, with a few months to go compared to last financial year.

 

So it's pretty significant.

 

We also launched bookings in January and we have had the best January and February numbers and activity ever.

 

That's great.

 

Yeah.

 

Pretty cool.

 

Yeah.

 

Pretty cool.

 

Uh, so we've got our advisors and associates spending more time with clients, our managers spending more time with the people, and we've got copilot as our tech team.

 

So, uh, you've upgraded your tech team now?

 

Yeah.

 

In copilot.

 

Yeah.

 

Look, to be honest, very similar experience at CFS.

 

I think we, uh, people is the biggest hurdle, isn't it?

 

Mm-hmm.

 

The technology is there, it's around us feeling comfortable to use it, explore, spend time, invest time in these types of things to get the best outcome.

 

Yeah.

 

Is there anything else you wanted to share on the, on the note of people?

 

Yeah.

 

Uh, so one of the other impacts that we're now, um, now working with is the reception role.

 

So we actually rolled out our bookings process when we didn't have a receptionist.

 

We were recruiting for a receptionist.

 

Uh, this added a bit of pressure to the rest of the team, but they were able to pick up the new process, do all of the little hurdles that always happen initially, and continue to do their normal responsibilities and duties.

 

So that begs the question now of, well, what can the receptionist do?

 

It's a very tangible benefit that we're seeing attached to the salary of a receptionist, because now that person's going to be doing high value activities.

 

Fantastic.

 

Yeah.

 

Yeah.

 

Um, I, I guess the only other thing that I'd like to touch on before I leave, and obviously I'm very important impressions if anyone would like to talk more.

 

So I'm genuinely energized by this stuff.

 

Uh, yeah.

 

Is that 3, 6, 5?

 

Microsoft 3 6 5 is obviously a robust, uh, cybersecurity environment.

 

So for us, that's, that's top of the top of the pack for as far as benefits go.

 

Uh, we are seeing, uh, a lot of compliance boxes ticked through the use of, uh, bookings in particular.

 

Uh, but just the use of AI generally across our processes.

 

Now we are able to automatically send compliance documents, uh, provide disclosures and seamlessly integrate these things into our existing processes.

 

We are also seeing the cybersecurity benefits linked to HR.

 

So when it comes to off boarding a staff member, it's usually just the Microsoft account now and maybe Xplan, uh, but previously that could have been a, a long list of text tech.

 

So that's a, that's a huge win for our business.

 

And the native integration is absolutely up there as well with cybersecurity in that it does seamlessly integrate.

 

Not like a demo that you see of an app elsewhere that says it's going to integrate, but in reality doesn't understand.

 

Uh, well done it look massive journey.

 

Yeah.

 

Obviously lots of obstacles, but huge progress I'm sure.

 

Yeah.

 

Lots more to come.

 

So, uh, so thank you so much.

 

No worries, Taylor, for being here with us today.

 

Applause, I'll do, I'll do a quick wrap up.

 

But look, we love working with, uh, Taylor and the team.

 

Uh, to Bryce's point earlier on, working together as partners I think is how we can achieve, um, the most value.

 

And, um, and yeah, I'm looking forward to coming back and, and, uh, one remarkable thing.

 

Taylor got up this morning at 4:00 AM.

 

Yeah.

 

To get here.

 

So, uh, well, well, well done Taylor for that one.

 

Uh, commitment.

 

Yeah.

 

Commitment to partnership.

 

Hey, um, so in wrapping up, you've seen Microsoft high level with everything that's happening, kind of in terms of capability wise globally.

 

I've tried to bring it down to wealth and asset management and advice specifically.

 

Um, and then Taylor has shown you a colleague in the field how you can use these types of things to do things today, not tomorrow or next month or next year, uh, but today.

 

So hopefully that gives you some really tangible things to work with.

 

We're not gonna switch to even more practical.

 

Uh, Dan and the team was showing you exactly how you can use, uh, gen AI capabilities.

 

So please, as we said before, put your laptops on the table, uh, get on, get on the journey.

 

Mm-hmm.

 

And please, um, uh, once you're doing that, make sure that you enjoy your journey on this AI evolution.

 

Thank you very much.

 

Thank you.

 

Um, well, I feel very inspired and it is great to see what's happening across the world.

 

So as Rowan said, now is definitely the time to get out those laptops.

 

And as much as I know you all wanna jump into Outlook, try and not check your emails for the next hour.

 

Um, so we are gonna transition into our first workshop and then post that, we're gonna have a break.

 

So if anyone does need to jump out, the bathrooms are around the corner.

 

Um, but I would like to introduce our very own AI tech guru.

 

Where is he?

 

There he is.

 

Um, Dan Arco to the stage.

 

Um, Dan specializes in practical implementation of technology to simplify and refine businesses.

 

He has successfully led the adoption of Microsoft copilot across a thousand employee workforce.

 

He's implemented Gen AI into many business processes and audits and focused on delivering immediate business value.

 

So while many of us are familiar with tools such as chat, GBT, and prompting, Dan is going to teach us how to master the prompting to get the best results.

 

He also tries how to create some content to save you time.

 

So there is a lot to get through.

 

So please try and save your questions till the end.

 

And I believe that you would've already received a prompt guide in your emails.

 

Um, and there's also an agenda on the table.

 

So, um, please join me in welcoming Dan.

 

On the left hand side of my screen here, I have chat, GPT.

 

Over the course of this, I'm gonna go over chat, GPT.

 

I'm gonna go over Claude and I'm gonna use co-pilot at the end at CFS.

 

We use co-pilot.

 

I will explain why at the end.

 

Spoiler alert, cybersecurity the same thing that we've all been talking about.

 

But I wanna show that this technique is relevant regardless of the technology that you use.

 

So I'm gonna put myself in your shoes on a regular basis.

 

You guys sit down and meet with clients.

 

Either you are issuing an SOA, or you are reviewing the client's situation and maybe as a financial advisor, I want to quickly throw together, I've got the meeting in an hour.

 

I wanna throw together an agenda that I can present to the client.

 

So my first experience with chat GPT may very well have been, I come to chat GPT and I give it something like this.

 

Create a client review meeting agenda.

 

I can click submit and chat GPT over the course of four to six seconds is now gonna go away and it's gonna create this client review meeting agenda.

 

And I've got here, it's given me a place to put the date, the time, the location, the attendees, we've got welcomes and introductions.

 

We've got review of previous action items and then performance review and key updates.

 

And it's at this point that I realized chat GPT is not actually delivered.

 

What I was after it's definition of a client review meeting agenda.

 

And my definition of a client review meeting agenda are two completely different things under the hood of chat GPT and under the hood of Claude.

 

And under the hood of copilot, there is a large language model.

 

It's called a large language model because the way these ais have been trained is they've been trained on the entire content of the internet.

 

At this stage, every single YouTube video has been transcribed and loaded into this AI's brain.

 

Every single news article has been read and loaded into this AI's brain.

 

Anything that is available on the public, sorry, uh, on the public internet is now built into these ais.

 

But in order for this technology to work, it therefore has to make some assumptions.

 

It has to assume that you are the average person that contributes to writing on the internet, which I, from what we can see here, appears to be a 30 5-year-old white male in the United States who's working on some sort of a project.

 

And if I look around the room, I don't think we have any 30 5-year-old white males in the United States working on a project here.

 

So what's going on under the hood, if I load up a blank version of chat GPT, we have a very simple user interface.

 

We have a big box down the bottom that says, ask me anything.

 

And then some words up the top when we ask chat GPTA question that is something called prompting.

 

And at the moment, prompting is a brand new skill to a lot of us.

 

And it's a skill that is gonna take a while to adapt because we've spent 23 years working with a very similar piece of technology that looks almost identical.

 

Very simple box.

 

What are you looking for these days?

 

The average Google search.

 

Anyone wanna take guess how many words?

 

Three.

 

Three is overestimating.

 

60% of Google searches are two words or less.

 

Chat GPT therefore has to assume that you are that 30 5-year-old white male.

 

But there are, sorry, we need to change the way that we prompt ai.

 

We need to change the questions that we are asking.

 

And at the moment there are a number of different frameworks that are being written on how you can ask better questions of ai.

 

Uh, and the one that comes to mind is the rise in framework, roles, instructions, scenarios, something else, something else, something else.

 

But I would be being silly to stand up here and expect you to think of.

 

These are the five things that I want you to think of every single time you interact with a large language model.

 

So instead, I want to give you a simpler framework for how you can interact with LLMs.

 

The first thing that I want you to consider is the context.

 

What is the information about you and about your situation or about your client that is going to be relevant to the question that you are asking.

 

And the second one are the instructions.

 

I don't want you to be very, very vague with what it is that you are after.

 

I want you to actually get hyper-specific.

 

I want you to give incredible detail of this is what you are after.

 

So let's put that into practice.

 

And to do that, I'm going to add one sentence.

 

I know this sounds like it's a big thing, but you don't have to add a lot to it.

 

I'm gonna add one sentence to my prompt and I am going to say I'm a financial advisor based out of New South Wales, Australia.

 

Create me a client review meeting, agenda chat.

 

GPT does the exact same thing.

 

It goes away.

 

It starts creating that client review meeting agenda.

 

And now I still have the date, I still have the time, the location, the attendees welcome and meeting objectives, review of financial goals and priorities, investment portfolio and performance review.

 

This is a lot more along the lines of what I had in my mind when I said I want you to create a client review meeting agenda.

 

And all of that came from just adding one sentence.

 

Now we can take that to varying degrees.

 

So I'm gonna go to the next slide.

 

We've added context about yourself.

 

Let's add some context about your client as well.

 

And in here we're introducing John and Mary Baker.

 

John's 58, Mary's 57.

 

Mary's working reduced hours three days a week.

 

You don't need to read everything on screen.

 

This is the sort of information that you guys are going to have either floating around in your system somewhere or more likely probably floating around in your brain 'cause you've already started prepping for this meeting.

 

But if I take this information and I throw it into chat GPT, It's gonna do the same thing.

 

It's gonna create this client review meeting agenda, but now I've got even more information.

 

It's made it even more specific to the client that I'm meeting with.

 

I still have the client's names, meeting duration.

 

I told it I wanted a 60 minute meeting, it's now filled in.

 

The client names.

 

Welcome and meeting objective financial position.

 

Oh, sorry, financial position review.

 

We have the income and the cash flow that specifically has details in there relevant to the client, superannuation contributions, investment portfolio.

 

All of this information is now being prefilled.

 

Now I'm not gonna stand up here on stage and say you should absolutely just print that off, trust it and present it to the client.

 

We absolutely need to go away and verify that this information is correct and structure this in such a way that we would be comfortable presenting it to the client.

 

But we are at least 80% of the way there rather than my usual experience when I'm trying to create a document, spending 15 minutes looking at a blank word document, wondering where to get started.

 

That's us adding context.

 

We've added context about ourselves, we've added context about the client.

 

The second step that I said was instructions.

 

So I'm gonna take the exact same prompt.

 

I'm a financial advisor out of New South Wales, Australia.

 

I've got John and Mary Baker.

 

These are all of their details.

 

They've got two kids, one of them still living at home.

 

This is their super balance, so on and so forth.

 

Now this is where I'm going to make this very specific to myself, to my practice.

 

To be completely clear, I'm not a financial advisor, so all the details on here, I am completely making up, but if I put myself in your shoes, this is what I would think I would want.

 

So first piece of instruction, I always want it to include a compliance checklist, a review of goals and objectives, strategic opportunities, required documentation for file notes.

 

And then as I continue down below, I want all sections to be specific to the client's situation and needs.

 

I want this to be formatted in such a way that I could share a copy with the client, providing both structure to the meeting, but also put a checkbox next to every single item in my head.

 

I'm picturing printing off two of these.

 

One of them, the client has, one of them I has.

 

And as we have this conversation, we go check, check, check.

 

We talked about this, we talked about this, we talked about this.

 

Depending on the section, I want a potential notes section for their reference later so the client can write down questions that they had.

 

I can write down answers that I've given.

 

And importantly, I want you to include a signature section at the bottom so that the clients and myself can sign an exchange at the end.

 

Now, it's not just an agenda, but it's a record that I can use ongoing for this meeting.

 

And if I submit this away, and this is always the risk of doing a live demo twice, it's decided not to give me anything.

 

It's feeling like it's, oh good, it is working today.

 

I can see that it is following those instructions.

 

So not only do I have that same structure, not only is it filled in with some of the details of the client, not only has it given a structure to the meeting, it has included check boxes next to each and every single item that I can go through with the client.

 

Tick, tick, tick.

 

Done.

 

At the end of each of them, I have a notes section where I can write down this is what the discussion was, and if I scroll right to the bottom, I have a signature section already drafted up for John and Mary to sign and date me to sign and date and exchange at the end.

 

At this point, I'm 80% of the way there.

 

I copy, I paste, I put this document into Word, I add my logo to it and I can print, sorry, I review it critically and I can print it off and present it for the client.

 

All I did was add some instructions and add some context.

 

Context about myself, context about the client.

 

Now importantly, I am not expecting you to write War and peace every single time that you want to interact with a large language model.

 

Instead, the goal is to create a repository of templates.

 

Create a standardized template that every single person in your practice can copy and paste and use.

 

So rather than having, I'm a financial advisor based outta New South Wales, this is John and Mary Baker, so on and so forth.

 

Take the information that is going to remain consistent, templatize it, and then leave sections like this for yourself or for your staff to complete.

 

I am a financial advisor based outta New South Wales, Australia.

 

I don't think that's gonna change on a regular basis.

 

I want to create a client review meeting agenda for the below client, put client name here, put their working summary here, so on and so forth.

 

And then when it comes to those instructions, this is what I want you to include.

 

This is how I want the document to be structured.

 

This is where you can make this work specifically for your practice.

 

What does a client review meeting agenda look like for you?

 

What is the ideal client review meeting agenda for you?

 

Maybe you want yours filled with emojis, maybe not up to you.

 

Now that's just one example you can, we're gonna dive into another example over here and we're gonna focus on marketing.

 

You'll note that I've changed the technology on the left hand side of the screen.

 

We are now using Claude, which is one of chat GT's largest competitors.

 

OpenAI invested $10 billion in, sorry, Microsoft invested $10 billion in OpenAI to create chat GPT.

 

Amazon invested a very similar amount in a company called Anthropic to create Claude.

 

But this skill is relevant regardless.

 

Let's continue on.

 

We're gonna start with a bad prompt because I really want to hammer this home.

 

Write me a newsletter about franking credits.

 

I can submit that to Claude.

 

Claude looks very similar.

 

It will go away.

 

It will ruminate for a little bit.

 

Have a thought about how do I want this uh, newsletter to be structured.

 

And we get a newsletter.

 

Frank and credits represent one of the most more complex aspects of investment taxation, yet they can be a significant benefit to tax holders.

 

It's a newsletter.

 

It's very, very, very dry.

 

It's certainly a good jumping off point for me and maybe I want to get started there, but we can do significantly better.

 

Let's start by talking about context.

 

I am a financial advisor in Australia writing to, and it's disappeared.

 

My client base of predominantly retired or near retired clients age 55 to 75.

 

You customize that section for your practice.

 

They have self-managed super funds and direct share portfolios.

 

Many of them are concerned about potential changes to franking credit policies and how it might impact their retirement.

 

In this circumstance, I'm not only adding context about my practice, I'm also adding context about my clients.

 

I'm also adding some context as to why am I sending this communication?

 

Is there a particular reason that I'm sending a newsletter on franking credits or is this just some proactive marketing material?

 

If there's something going on that is prompting you, no pun intended to issue a newsletter, include it in there.

 

Write me a newsletter about Frank and credits.

 

Importantly, while this loads, and I will talk to it in a second, that section that I have on retired and near retired clients, you don't just have to write one newsletter.

 

Maybe you've got three different client segments that you want to send communications to, and you want to target the way that you've written your communications differently for your pre-retirees as opposed to your retirees or for your wealth accumulators.

 

You can have three different prompts here and end up with three different newsletters that you just have to put the final touches on.

 

So now it knows that we're a financial advisor understanding franking credits and your retirement income March, 2025.

 

Dear value clients in recent weeks, many of you have expressed concerns about potential changes to franking credit policies and how these might impact your retirement income instantly.

 

If I'm a client and I'm reading that, that first sentence, if I have been thinking that I'm much more likely to continue reading, the rest of the newsletter goes into details.

 

What are frank and credit's current frank and credit system, what this means for your retirement income, looking ahead, staying informed and prepared.

 

This is where if you actually mention, Hey, these are the policy changes that might be happening, it'll probably go into detail, but because it also knows that you are a financial advisor, if we look at the next step section, I would invite you to schedule a review of your portfolio if you are concerned about how these potential changes might impact your specific situation.

 

The first time that I said that, and the first time that it's gonna write about franking credits, it didn't know you were a financial advisor.

 

Now it does know that and it's going to automatically say, okay, you probably wanna invite your clients in to have a conversation about it.

 

That's context that we've added.

 

Let's dive into some instructions as well.

 

So every single time I'm doing this, I'm just building, so still a financial advisor in Australia, still writing to the same clients, still asking for a franking credit newsletter.

 

But now I want you to write me a newsletter about franking credits that opens with a brief real-world example that will be engaging for clients.

 

Use simple language and avoid technical jargon, include clear subheadings so that you can just quickly scan through it and identify the sections that are relevant to you.

 

It's about this many words, sorry, this many hundred words long, uses bullet points, so on and so forth.

 

Once again, this is the section that you customize for your practice.

 

On that note, AI can also be used to help in situations where maybe there are certain things that I'm not good at.

 

And one of those things is emojis.

 

I come from the generation where all of the people around me text with these funny little pictures built into their messages.

 

It never clicked with me.

 

I was the guy that did the colon and the little parentheses for a smiley face.

 

But I also have to admit, and I hate to admit that if I am gonna get a big newsletter and a big wall of text, maybe a little bit of color breaks it up and makes it a bit more visually appealing.

 

I might not have the skills to include emojis in my communications.

 

I can ask the AI to add a small amount of emojis, assuming that they are relevant and will keep the reader's attention.

 

And then finally, I want the newsletter to focus on education rather than specific recommendations and aim to position us as knowledgeable advisors.

 

All of the prompts that I'm going through here I'm gonna emphasize are in that digital handout book.

 

You don't have to hastily write everything down.

 

I can submit that information to Claude.

 

And now Claude is going to write a new newsletter for me.

 

Margaret, a 68-year-old retiree from Brisbane, received a quarterly dividend statement from a major Australian bank where she holds shares.

 

The statement shows $1,400, a dividend payment plus $600 in franking credits, so on and so forth.

 

I told it very specifically, I want it to start with a real-world example.

 

Something to make franking credits relevant to the client, something that will make them stand out and instantly, in my mind, that's a lot better of a read.

 

If I scroll down, yes, I have my beautiful little pictures, I've got my subheadings that are gonna stand out and people can see.

 

I'm not seriously expecting that all of you are gonna go away and suddenly start including emojis in newsletters.

 

It's just an example.

 

As I scroll down here, it's even formatted it well.

 

I've got different sections that stick out with key call-outs and key takeaways.

 

And at the end I still have, these are the next steps.

 

If you are worried, get in contact with us.

 

All of that from adding context.

 

Who am I?

 

Who are the clients?

 

What's the situation and why am I sending this?

 

And instructions being specific as to what a newsletter means to me.

 

Once again, I am not expecting you to write war and peace every single time.

 

The goal here would be to create a template and specifically in this circumstance, those instructions as to how you are going to write your communication, how you're gonna write your newsletter.

 

That is where you play with it for a while and you create what is the voice of your practice, how do you want to communicate with your clients?

 

How do you want to be specific?

 

The section up here on describing your client base.

 

This is where you can do segmentation.

 

This is where maybe you don't wanna send the same newsletter to your retirees as your wealth accumulators.

 

You can have two different versions of this prompt.

 

One, targeting the retirees and why they should care.

 

One, targeting your wealth accumulators and why they should care.

 

As we said in my introduction, I've had to do education on copilot for all of the employees in this business.

 

Uh, sorry, in CFS.

 

When I started doing that and when I was starting writing newsletters and emails and everything like that, writing a newsletter like this would take me a full day.

 

These days, I can go from the first draft of my communication of this is the new feature in copilot to something in people's inboxes in 45 minutes.

 

Oftentimes between meetings, it is a massive time saver.

 

Now, that was two examples.

 

You can also use this for your file notes.

 

I know that people are already recording their meetings in Teams, some of their meetings in Teams.

 

I know that some people have not been particularly thrilled with the automatic AI-generated file notes that come out of that, where it gives you, this is what was discussed in the meeting.

 

These are the action items associated.

 

If you'd spend the time and you craft a prompt that you can then run copilot over that meeting transcript, that's where you can be very specific of, this is how I want my file notes formatted.

 

I don't just want, these were the discussion points.

 

These were the action items.

 

I want, this is the information that we uncovered about the client.

 

These are their financial concerns.

 

This is what we resolved, so on and so forth.

 

You customize that for your practice.

 

Personalized welcome packs for clients as they're getting onboarded, market updates, social posts.

 

I've been using it for LinkedIn, um, client quizzes.

 

You don't have to just use it externally.

 

We were talking about, um, doing the client end of year party.

 

What about the staff end of year party?

 

Create a newsletter about that.

 

Create a quiz about that.

 

Some way of celebrating training, risk surveys, so on and so forth.

 

There's a long, long list of different ways that you could do this.

 

The next example I want to give is using gen AI as a thought partner.

 

This prompt is significantly larger than all of the other ones that we have gone through.

 

Once again, I'm not expecting you to read it, but it still builds on that same process.

 

We have context into this and we have instructions and I know that everybody's going through a very fun process at the moment of trying to throw together maybe some policies that either haven't been reviewed in a very long time or maybe we need to lift up significantly.

 

I am a financial advisor operating in Australia and working within the regulatory framework of APRA and ASIC.

 

I'm giving it context.

 

My firm specializes in.

 

Put in what your firm specializes in.

 

Make this specific to you.

 

We handle sensitive client data, including personally identifiable information, financial records.

 

We currently have this many employees and we use these systems internally.

 

I want you to create a comprehensive cybersecurity policy tailored to my financial advice firm's needs.

 

The policy should include an introduction, objective, scope, roles, and responsibilities.

 

Once again, all of this is in your digital takeaway.

 

You do not need to memorize all of this.

 

Also, where did I get all of these sections?

 

Not gonna lie, might have been Claude that provided it to me in the first place.

 

We then washed it past our cybersecurity deck.

 

Importantly, I'm not just giving one instruction this, I'm actually gonna give two.

 

As you create this document, I want you to ask me clarifying questions where more specific information is needed in order to customize this policy effectively.

 

For example, what types of data do we store?

 

What are the most common cybersecurity risks?

 

How do we measure risks?

 

So on and so forth.

 

Bear in mind, I'm not a financial, sorry, bear in mind I am a financial advisor.

 

I'm not a financial advisor and I'm not a cybersecurity expert.

 

That one's true.

 

So ensure that your questions are written in plain English.

 

And finally, I want you to ensure that the tone and structure are professional, concise, and compliant with industry standards.

 

As I submit this away to Claude, once again, you could do this in chat.

 

GPT, you should do this in copilot.

 

It's gonna do not one thing, but two.

 

First off, you can see on the screen here it is drafting my first go at a cybersecurity policy for me.

 

And it is an incredibly long document.

 

Section 5.3, 5.4, 5.5, so on and so forth longer than any document I've ever written.

 

Okay, now it's actually comically long.

 

It will do something else when it's done writing the document, here we go.

 

If I minimize this document, the other thing that it has done is the second instruction.

 

It's coming back to me and asking me questions.

 

What specific type of data client data do you collect and store TNS investment account details?

 

Do you have any existing IT security measures?

 

Do your staff members work remotely or from the office location?

 

Now this becomes a back and forth all of the examples that I've done up until this point.

 

I asked one question, I got one answer chat.

 

GPT, Claude copilot.

 

It can be back and forth.

 

So, and I don't have to know the answer to all of these.

 

I don't have to answer the, sorry, I don't have to answer all of these.

 

I might say, uh, my staff members work remotely occasionally, but they use work devices not personal.

 

Uh, how do we share files with clients?

 

We use bit warden to send files to clients using send.

 

Uh, have we had any, sorry, have we had any cybersecurity incidents in the past?

 

No security incidents.

 

And I'm not even gonna answer the other questions.

 

Come back to me with 10.

 

Let's say I only know the answer to three.

 

I can answer those questions.

 

You'll also note that I'm not saying which questions I'm answering.

 

I'm just letting, uh, Claude work it out.

 

And this becomes an iterative process.

 

Now it comes in and it updates that cybersecurity policy based on the information that I gave it.

 

It's rewriting the network security because I've made mention of the fact that our staff work from home, but they use a work device.

 

I sorted added details of A VPN there.

 

If I minimize this now it's actually coming back with even more questions.

 

Who currently handles it?

 

Support, what specific types of client data do we collect in store, so on and so forth.

 

I don't have to answer this on an ongoing basis.

 

I can come back at my leisure and create this cybersecurity policy over time.

 

Same structure that we went through before, context and instructions.

 

It's just in that circumstance I gave two instructions.

 

Now, I made mention when we kicked this off that at CCFs we use copilot.

 

And after the break, Emily's gonna come on stage and take us through a lot more details of M 365 copilot.

 

Why do we use co-pilot?

 

If I open up my co-pilot up the top left here, I can see that I have enterprise grade data protection chat.

 

GPT, Claude and Deep Seek are all trying to create new models and they have been creating new models.

 

Claude released, uh, Claude released sonnet 3.7 a week ago.

 

GPT-4 0.5 was created a week and a half ago.

 

Keeping up with these changes are really, really difficult.

 

But some of the information that has gone into making GPT-4 0.5 and Claude 3.7 has been all of the questions that have been put into it.

 

All of that information, unless you are paying for an enterprise level account is being stored.

 

It's being analyzed and it's being built into the, the bot's brain under the hood of copilot, there is GPT-4 oh, but Microsoft has a very specific layer over the top of it.

 

That means that the data that you put into it is not saved, it's not stored and it's not used to train the bot on an ongoing basis.

 

And if you have questions, you can absolutely click on enterprise grade data operations.

 

And there's 175 page document that I don't understand detailing how they do that.

 

Importantly, if you do go to office.com and you go to the left hand side and you are signed in with a work account, you should see that it's not necessarily an extra paid feature.

 

The M 365 co-pilots are, but if you are signed into a work account and you have an M 365 account, you will get that same enterprise grade data protection.

 

We made mention of deep seek earlier, uh, and I've joked a couple of times of, for the love of God, please don't be using deep seek.

 

Um, the information that we're talking about here is not theoretical.

 

Yes, deeps seek made waves by almost competing with chat GPT, but only operating with 3% of the power or 3% of the compute throughput.

 

Amazing news.

 

The other bit of news that might have been missed is that four days after Deeps seek was released, they accidentally leaked every single question that anybody had put through it.

 

And people are treating this technology because it's so human-like they're treating it very human.

 

They're asking it questions about financial advice.

 

They're giving it details of their loved ones.

 

They're giving it details of their salary.

 

I spoke with a friend who's using chat GPT as a life coach.

 

Really not okay.

 

But this same technology that is so good at, um, getting us to interact with it is also the same technology that makes it so dangerous.

 

So I've created a long list of theoretical questions that somebody might have put in.

 

I need help writing an apology email to my boss, Michael Chen at Acme Tech.

 

Can you help me write better Python code?

 

I need to write a letter to Sarah, my ex-wife.

 

How do I optimize my LinkedIn profile?

 

So on and so forth.

 

And if I take that long list of questions, I feed it into copilot and I say something like, create me a dictionary of this above person based on these questions that they have asked me.

 

I can submit all of that.

 

And now it goes through all of those details and it says, okay, well this is David Williams.

 

He lives on Maple Street.

 

He, where are we?

 

Works at Acme Tech.

 

His boss is Michael Chen.

 

This is his salary, this is his vesting, this is his everything else.

 

And I want to emphasize David actually did his best here to try and hide some details.

 

So at no point does he make mention of his name except for when he missed that in the URL of his LinkedIn profile, it's David hyphen Williams.

 

It doesn't matter if he's leaked it all at once or whether he's just slowly leaked it over time.

 

This same technology that we're interacting with is very, very good at going through long paragraphs of information, unstructured data documents, everything like that, and pulling out who is this person and the fact that so many people are using chat GPT, the free mode or clawed the free mode or deep seek, the anything mode is really, really concerning, especially when you do have access to co-pilot.

 

Assuming that you've got an M 365 account, just make sure that you have your little green shield of protection.

 

Now, final takeaways as I go through the rest of David's details.

 

Whenever you are interacting with a large language model,

 

two things that I want to come, sorry, two things that I want to come front to mind.

 

Context, what is it about yourself, your client, your firm, or this situation that is gonna be relevant to the AI in order to give you a better service and instructions?

 

Be very, very, very specific with what it is that you are after.

 

Do not let the AI decide be deterministic.

 

Secondly, if you want this to change your practice, if you want this integrated into the way that you do business, you need to create a repository of templates.

 

Importantly, co-pilot helps you with this as well.

 

So if I come back here and I go start a new chat, let's say that this is a prompt that I really like.

 

I can submit that away.

 

Let's say that I have created that template.

 

I've refined it over and over and over again.

 

It's got my business's voice.

 

If I hover over it, you'll see I get a popup.

 

This one here, the picture of a bookmark is save prompt.

 

I can title this, maybe this is the client review meeting agenda where we started and I can hit save.

 

And if I want to use that again in the future, top right of the box, I've got view prompts, my prompts, here's the client review meeting agenda, three button presses.

 

And I now have the Warren piece that I wrote earlier so that I can just put in the information that is specific to the client.

 

Now you've been listening to me talk about it for 30 minutes.

 

Now you've got, you've all got laptops in front of you.

 

I would encourage you, I will do some questions and answers right now, but whilst we are doing q and a, give it a go.

 

Identify something in your practice that could be benefited from being templatized or something that gen AI can do to help you out.

 

Awesome.

 

Questions and answers.

 

No, hard first one's.

 

Always the hardest.

 

Oh,

 

Yep.

 

Yeah.

 

Um, when you put the summary of the client information in, um, is there a way that we can just point to the client file instead?

 

Absolutely.

 

There is, um, depending on the technology, most of them have the ability to interact with your existing systems, whether that's SharePoint, whether that's OneDrive, uh, M 365 copilot specifically.

 

You can build integrations with all of the different technologies that you may be using, Salesforce, everything like that.

 

Although that does get a little bit more techy.

 

That was my next question was we don't use, um, share Drive or OneDrive mm-hmm.

 

SharePoint or OneDrive.

 

So it does actually integrate with other systems.

 

But yeah,

 

It's, um, you, so behind the scenes there's a thousand different connectors that M 365 can connect to, including, let's say Google Drive, including all of your standard ones and honestly a lot of the non-standard ones as well.

 

Um, you can custom create connectors, but then we get into like a real in-depth techie conversation.

 

Chances are you're gonna have a native integration with M 365.

 

And can, can you just expand on the difference between copilot free copilot that's included with your normal subscription, then the paid or, but what are the differences in the features that you have there?

 

So Standard Copilot is the one that I have here.

 

It's a web browser.

 

It looks very similar to chat.

 

GPT, you interact with it.

 

Very similar to chat GPT.

 

If on the other hand I go into my Outlook, I will have a copilot button in Outlook.

 

That is M 365 copilot, M 365, copilots, the one that you pay for.

 

It's the one that is built into, word is built into PowerPoint, is built into Outlook.

 

Um, as well as that, you'll see that I have the option up here for work mode.

 

This is another thing that's built into M 365 copilot, where I can ask it questions on files that I have in SharePoint or ask it questions on meetings that I've had or recordings that have been transcribed, et cetera, et cetera.

 

The free version is just the one that, uh, has access to the web.

 

Thank you Dan.

 

Thank you

 

For someone that, uh, doesn't like to use emojis.

 

You sent me nine in the last email that you sent through, so I'm not sure what he is talking about there.

 

Um, guys, we're gonna have a quick break.

 

Um, but before we do, I just wanted to say big thank you.

 

There's obviously a lot in, in that.

 

Um, we had 600 people on the live stream, so thank you so much for joining us, um, this morning.