Presentation

Transcript

Title: AI as a Business Model | Marc Wilson, Global Advisors | techfest 2026

Date: June 17, 2026

Speaker: Marc Wilson, Managing Partner, Global Advisors

Introduction by Host Tony van Niekerk

Tony van Niekerk: Marc Wilson, our first speaker, gave us a good talk at our Innovators Network at the beginning of the year. He positioned the strategic importance of technology for us, looking at how business models are changing, and it fitted in perfectly with what we wanted to do today with TechFest. That is to look at technology as the catalyst for better insurance business, not just as technology for technology's sake. Marc is the founder and CEO of Global Advisors. He will explore where the biggest opportunities lie, talking about the fundamentals in terms of redesigning business models, customer relationships, and competitive advantage. Marc, thank you very much for taking the time and being available to discuss this with us. Without further ado, I will hand over to Marc Wilson, founder and CEO of Global Advisors. Thanks, Marc.

Opening Remarks

Marc Wilson: Let me just give a brief introduction. I'm Marc Wilson, Managing Partner of Global Advisors. I founded Global Advisors 20 years ago. Previously, I was at Gemini Consulting, where I headed up the financial services practice, and before that, I was at Deloitte. I have quite a lot of financial services experience across banking, corporate investment banking, retail, and life insurance. Most recently, I have focused heavily on artificial intelligence.

When I was at Gemini Consulting, I headed up the e-business unit. I lived through one of these cycles before. Some might call it a hype cycle, some might call it a .com bubble, or a bubble now. I can contrast, certainly for myself, what I'm seeing right now in AI with what I saw at the time in the .com bubble. That is quite useful, certainly for me, because I feel like I'm able to read some of the signals quite well. There are similarities, and there are differences.

One of the biggest similarities I see is that as we were leading up to the peak of .com, people talked about how it was going to put traditional incumbents out of business. Anybody on the call who was around in the early 2000s probably recognizes some of the existentialist type talk about what AI holds for the future of business and the future of insurers.

However, the difference, having lived through that and having led the e-business unit, is that this time feels different. There's a level of reality attached to this which there wasn't during the .com cycle. With the benefit of hindsight, we can see that industries have fundamentally changed since the .com bubble. Look at the media industry: it is unrecognizable from where it was before the .com bubble. Very few of you buy physical magazines and physical books anymore; that industry has changed. I'm sure many of you watch streaming video now. This time around, we probably need to look forward to see how this AI cycle will change insurance.

Section 1: The Threatened Customer Relationship

Slide 2 Insert: The threat is not a better chatbot

• Autonomous Agent Quote: "You're paying too much on this policy. I've run the comparison. Shall I negotiate - or move you?"
• Core Concept: The relationship with your customer gets intermediated by software you don't own.
• Precedent: Banking is already here - agents sweeping idle "lazy cash" balances overnight. Insurance is next.

Instead of focusing on a better chatbot, imagine an autonomous customer agent that says, "I've looked at your insurance. You're paying too much on this policy. I've run the comparisons. Shall I negotiate or move you?" For some of you, perhaps that sounds far-fetched. For others, perhaps you've already installed an agent like OpenClear or Hermes, sent it out to review your policies, and had it make a proposition to you.

Pete Steinberger, the developer of OpenClear, famously shared an example where he came back after lunch and his agent had negotiated a deal on a second-hand car purchase. That world is here right now. The technology is there, even if we aren't fully there from a business perspective. This world is technically completely possible today.

In my conversations with insurers—including those in technical spaces like commercial or fire insurance—these contracts are hugely bespoke and heavily negotiated. To them, this might seem completely far-fetched. But in reality, it means that, at the very least, the customer potentially has expert advice to review a contract and get feedback in real-time. It might not be doing comparison shopping for you initially, but your agent is able to review your insurance position, make suggestions, and help you negotiate.

The key question is: do you own that software, or has it been intermediated away from you by somebody else who owns the software sitting between you and your customer? Banking is already here; agents look at cash balances and move lazy cash overnight to get the best rates. They operate for the customer in a best-practice mechanism. Insurance is definitely on the way there.

Section 2: Core Strategic Questions

Slide 3 Insert: Three questions to hold for the next 40 minutes

1. Where is your value actually created?
2. Who owns your customer relationship?
3. What is your moat - what can't be bought off the shelf?

There are three questions I'd like you to think about during this talk:

• Where is your value actually created? It's a two-sided question: Where is it created for you as an insurer or broker—what justifies your commission and profit margins?—and where is it created for the customer? Why are they choosing you?
• Who owns your customer relationship? If you're a broker or intermediary, is it really you, is it the insurer, or is it the AI agent?
• What is your moat? What are you doing that cannot just be bought off the shelf?

This is top of mind for many right now. After experiencing Claude Fable last week, people saw a different level of AI and were switching between AI providers overnight. If that technology is generally available, anyone can pick it up and offer it to customers. A standalone model is never going to be the moat, especially if you are just putting a thin veneer of a package on top of it. Building differentiation on top of that is one of the most challenging questions relating to AI going forward.

Section 3: The Reframe & Our Journey

Slide 4 Insert: The reframe

• AI is not a tool you adopt. It's a layer forming underneath you.

Slide 5 Insert: Global Advisors, rebuilt as an AI-native firm

• Metrics: 600+ models, 200k+ RAG documents, 20bn+ tokens used, 100% employee agent coverage, 100% AI use 5 days a week.

This is not about a tool you adopt for your underwriters or customers. This is a layer forming underneath people's feet.

We have been active with this for the last two years intensively because I felt consulting would never be the same. I saw AI becoming an inference step in our traditional processes, taking over more of the core intelligence. We have had to fundamentally rebuild our business: our team has over 600 models available to them, we maintain over 200,000 RAG documents, and we've processed well over 20 billion tokens. Every one of our employees uses AI five days a week. We've made that journey toward being an AI-native firm. What is hidden behind those stats is that human behavior takes a long time to alter. Some of our team members are far better at getting results from AI than others, which tells us a large component of this change resides in the human element.

Section 4: Redesigning the Insurance Business Model

Slide 7 Insert: Insurance is moving from workflow automation to business-model redesign

• Not just efficiency: AI is becoming a new economic and decision layer that changes where value is created, captured, and who owns the relationship.
• Inference changes business models: AI inserts machine inference into deciding, explaining, prioritizing, qualifying, negotiating, and serving.
• The danger is mis-framing: The real issue is competitive redesign, not an operational technology rollout.

If you think this is just about workflow automation, you are undercalling the scope of the change. In the early 2000s, I remember large life insurers using electronic workflow to track policy applications across branches to manage OCR processing. Automation has been around for a long time. Today, electronic apps handle that automatically.

During the reengineering revolution, Michael Hammer and James Champy talked about "don't automate, obliterate." Reengineering had similar statistics to what we see today, with about 70% of projects failing. Today, reports suggest 70% to 90% of organizations adopting AI are not seeing the expected benefits. This takes time to bake into the efficiency layer, but you're mis-framing the problem if you're only looking at efficiency.

Inference changes business models. Automation was about workflows and deterministic processes—steps you codified and put through systems. That changes completely when you introduce automated intelligence and decisions. Expert systems have existed for a while, but not to this scope. This is about human-level or expert-level intelligence inside your workflows, allowing you to rethink your business model. This pressure to change business models will rise faster than most incumbents' ability to absorb it, and humans will battle to keep up with the pace of technology.

Section 5: The Competitive Timing Gap

Slide 8 Insert: Early movers that scale AI-led business model change first may create a lead late adopters struggle to close

• Timing Gaps matter (X vs. Y): Once peers scale, they define new ways to price, underwrite, distribute, and manage risk.
• Disruption Risk: Disruption risk becomes increasingly greater when AI fast movers transition from experimentation to scale at pace.

If you get a compounding effect but you are behind, you will always be behind, and that gap will get exponentially bigger over time. No matter how fast you catch up, you will never make up the difference. You know how this works with retirement savings: if you start saving at the end of your working career, you can't catch up to the person who saved the whole way through. The same compounding logic applies to AI inside a business. Competitors are trying to leap ahead and lock in a lead because they know they won't be caught.

Section 6: Frontiers of Disruption

Slide 9 Insert: The greatest risk will come from competitors that use AI to reimagine business models, redesign workflows and reset insurance economics

• Three Fronts: Enterprise readiness, Value-chain reinvention, and Business model disruption.
• Case Studies: Discovery (Behavior change via Vitality), AIG (Redesigned underwriting via AI ontology), and AXIS (Enterprise readiness and cultural reframing).

Slide 11 Insert: AI will reward brokers and insurers who own or enable the trusted intelligence layer

• Precedent: Aviva opened a ChatGPT-based quoting route; Visa and Mastercard are moving agent-initiated payments to live production.
• Strategic Choice: Own the trusted intelligence layer, enable the broker's layer, or drift into commodity balance sheet.

Slide 12 Insert: From payer to prevention engine - value moves upstream into prediction and behaviour

• Discovery Case Study: Shifted from passive risk pooling to behavior change. Recent Vitality AI partnership with Google moves tracking toward personalized, AI-enabled intervention.

We are moving beyond being claim payers and policy administrators to designing intelligent risk participation. This is not about tools; it's a layer forming underneath our feet.

AI is going to reward brokers and insurers who own or enable the trusted intelligence layer. Fundamental to insurance is trust. When you trust your money to a company, you expect them to be there when you claim, and to behave the way you expect. You cannot afford for that to go wrong. Embedding trust in the experience and the brand is key.

You probably won't see a ChatGPT-branded insurance product, but Aviva has already enabled an app within ChatGPT so you can have a discussion with Aviva through your AI chatbot. This lends the Aviva brand and experience to the comfort of a known relationship you have with OpenAI. I think that's where you have to look at how customers will behave regarding their need for trust and their desired experience.

The danger is that platforms like ChatGPT will host many apps, not just Aviva's, and might eventually say, "This is the best policy at the best price for you." The next version of ChatGPT is aiming to be a 'one app' for your online lifestyle. If that happens, insurance providers become like the Play Store or Apple App Store, installed in the back-end to give you access to the consumer.

The Aviva integration is important because it shows what can be done. Aviva is explicit that existing routes will continue to be there. When call centers first emerged, they tried to keep them as similar as possible to offline relationships because customers wanted that, and intermediaries needed assurance they weren't being undercut. Cannibalization could damage the broker relationship, and a very big channel conflict journey has to be navigated here.

Visa and Mastercard have also announced agent-initiated payments and commerce starting this year, showing that agents can safely shop and pay for you. If payments can be delegated safely, insurance renewals and endorsements can be delegated as well. Product superiority matters less if you don't participate well in machine-mediated search. Distribution power will shift to whoever owns the trusted intelligence layer.

Section 7: From Claims Payer to Prevention Engine

Value is shifting from paying and processing toward predicting, preventing, triaging, and orchestrating. The South African insurance market has already experienced this type of business-model disruption from within. Discovery's Vitality model demonstrated how behavioral data, incentives, and engagement can improve retention, risk, and economics. Its recent Vitality AI partnership with Google shows that the model is moving from behavior tracking toward more personalized, AI-enabled intervention.

The lesson is not that Discovery "used technology well"—it is that they changed the basis of competition by linking incentives, behavior, risk, and retention in a shared-value model. Highly engaged members have materially lower healthcare costs, lower admission rates, and lower mortality. Discovery's investment in Vitality AI aims to deliver more precise and personalized health improvement actions at scale, moving from rewarding behavior toward predicting where intervention is needed earlier. Somewhere in an insurer or an insurtech right now, there are people with an idea of how to do this differently using AI. The winners will be those who decide where trusted intelligence should sit, build the operating model to support it, and preserve scarce human capabilities.

Section 8: Summary of the Board Agenda

Slide 14 Insert: The winning question is not who has the best model, but who can deploy the most valuable inference in the most trusted workflows...

• The New Moat: Access to models is not a moat. Durable advantage shifts to context, workflow integration, enterprise memory, rights management, governance, distribution, and trust.

We think there are several things that should be on the board agenda for insurers, underwriters, reinsurers, and brokers:

1. Determine where intelligence should sit. When making decisions about prioritization, where do you get the most return on this intelligence? Is it customer-facing, broker-facing, operational, or the invisible infrastructure?
2. Map your business into three zones: Deterministic control (don't outsource reproducible workflows to inferential AI), inferential augmentation, and critical human judgment.
3. Build the harness. Recent research into leaked code shows that only 1.6% of the code is the actual model; the remaining 98.4% is the harness and scaffolding around it. What is the harness you are building inside your business around this inferential capability?
4. Choose your distribution posture. Where is the trusted intelligence, and how do you enable it for your partners?
5. Stop managing AI as a fragmented innovation portfolio. Start thinking about 3 to 5 design choices for your business as a whole.
6. Redesign capability and apprenticeship so the organization can absorb the change. This is about making sure all people in your organization become accomplished at working with AI on a day-to-day basis.
7. Prioritize local context. In the African context, make language, local data, infrastructure, and regulatory trust first-class parts of the design. Language is a big issue; most AI out there is not local language capable in the African context. If someone gets that right, they could deploy new AI that allows consumers to speak in their vernacular to the AI experience.

Ultimately, what kind of insurance are we trying to become before somebody else decides it for you? As I said in our last talk, AI is an amplifier. Success is not going to be knowledge; it's going to be your character—your organization's character in terms of how you respond.

Thank you very much.

Q&A and Closing Discussion

Tony van Niekerk: Brilliant, Marc. Thank you so much. I knew you would ace that one. Thanks for the fascinating information.

I think the one thing that I would want to ask you—and I think it might be on most people's thoughts here—I go down rabbit holes with regards to anything, and AI obviously has been one of those rabbit holes. The insurance industry's reaction to it has been... I've spoken to hundreds of people about it. I've had the pleasure of listening to you twice now and reading quite a bit of your stuff on your website, etc. My question is, where do the average business people, like myself and many of the brokers that are on here, where do they go for help? Where do they go to find out what should I be doing? How do I use all of this stuff that you've been talking about? How should I think? It's good to read like you set it out very nicely, but who's going to hold my hand in this journey?

Marc Wilson: Well, thank you for that step, Tony. Global Advisors, of course! (Laughter)
I'm kidding.

Look, I think that one of the AI-native companies out there—I think it was Every—said that people at the CXO level need to be spending between 10% and 20% of their time every day on AI. My experience is that no CEO has got that kind of time, right? So, it's a really difficult thing. The pace of moving with this is intense; for the last two years, I can personally say I have hardly slept. Consulting is completely different going forward from what it was, and we are dealing with a revolution which really has demanded every fiber of my being to get our business ready for that. It changes so quickly. On Wednesday last week, we had Claude Fable, and on Saturday morning, we didn't have Claude Fable. This stuff moves so quickly. My view of what AI could do was very different on the Wednesday from what it was on the Thursday. It's just incredible how fast it is.

While I said that piloting and experimentation are not enough to get you there on their own, it is absolutely necessary that you do pilot and that you do experiment, because it's only when you really push the envelope with trying different things and getting familiar that you can actually see what this stuff can do.

At a personal level, what I would suggest is that there are some people out there who are putting out amazing information every day about what's happening, and I can't cope without them. One of them is a guy called Nate B. Jones; he literally has an hour-long session every day where he's going through what's happened and what the models can do. That's an hour every day, so you have to pick these things. Something more accessible is probably a guy called Matthew Berman, who talks about some of these things. Find a YouTube channel and just get familiar with some of the people that are talking about this on a regular basis. Matt Wolfe is another guy. There are these guys that are really good at talking about the latest developments.

I think you need to digest news in a microformat—use the old RSS feeds and X feeds to get the headlines and keep on touch of the daily happening. TikTok is probably something for people who are more comfortable with that; the algorithm sends you stuff, so the more you start inquiring about this stuff, the more your algorithm will change to show you more of that content. That's at a personal level.

At an organizational level, there are two things that we are seeing which are fundamental. During the internet age, I ran a corporate incubator where you could take projects out of the organization, put them in the incubator, and give them all of the fuel and assistance necessary to succeed without traditional organizational constraints, and then deploy them back into the organization after they had found their feet. Many of the biggest insurers in the country have used skunkworks operations. I once had to go out of a building, walk around to a side door, and go into a basement where a team was working on the next version of that insurer. They had to take the people away from the noise and give them the space to think about the future. Unfortunately, almost all of those skunkworks programs failed; there are very few that have actually succeeded. But the incubation or acceleration example is definitely a format whose time has come again, where people need space to say, "We are dedicating focus to this."

Cycle people through that where you say, "One day a week, you are going to spend time in the incubator or the accelerator," or dedicate a set number of hours a day from the organization where they can see the art of the possible. We'll go live probably tomorrow with something on our website about AI visioning. The art of the possible is key because the biggest difficulty here is that most people just can't imagine what's coming. They can't imagine what's different, so they think in terms of their current role and their current business, while there are people coming from left field who are not constrained that way. Giving people a view of the art of the possible is a really big part of this as well.

Follow companies like Ping An, Aviva, and others, and see what they are doing. Look at the news articles coming out of those organizations, because if they are doing it, they are probably not as fast as the insurtechs and fin techs who are racing ahead without legacy constraints. You are seeing a lag indicator when Ping An or Aviva is doing it, as crazy as that sounds, but definitely subscribe to the news feeds around some of the big leaders.

Another big thing that I do is look at investor presentations. People tell you what they're going to do in investor presentations because they're talking up their share prices. Go and look at the investor presentations when they come out for the big players and see what they are promising for the current year. That's a really good insight as to what's coming down the track. Those are some ideas.

Another thing we've mentioned to one of the insurers is that I think shared services is a big part of this. Mercer, one of the biggest players in the world, has said, "We can't afford for all our different businesses to do their own thing here; they will all fail slowly and duplicate efforts." So they put together their own OpenAI, if you like, for the Mercer group, making fast-advancing AI available to players as diverse as their employee benefits advisory group and their consulting group. They are all drawing off a central AI capability that's accelerating the group as a whole. That is a really smart way of doing it to make sure that you are not diluting efforts or ending up with a stranded capability that everyone is not benefiting from.

Tony van Niekerk: Marc, thank you so much for that. I could carry on questioning you for a long time, but for the sake of this day, I need to move on. Thank you very much for sharing your knowledge, and we look forward to hearing much more from you. Make sure, guys, that you do visit Global Advisors' website because, as Marc said, there are lots of things there that are very interesting. He didn't mention that he is one of the people that you could actually follow and read about regarding what's going on in AI. Thank you, Marc.

Marc Wilson: Thanks, Tony. There are a lot of Global Advisors out there, so we are globaladvisors.biz with an 's'. You'll see that we put stuff out there, and often it's bite-size—quotes of the day, terms of the day. Usually, those have got a ton of information behind them, and you can get that through our social media feeds as well.

Tony van Niekerk: Brilliant. Thank you very much, Marc. Have a good one.

Marc Wilson: Thanks, Tony. Thanks, Brent.

Tony van Niekerk: Cheers.

(11:06 - 11:34 Electronic Outro Music)