Financial Services in the Agent Economy: From Platforms to Autonomous Operating Systems

Transcript

Let's try that Okay. Hello. Good morning CC. Uh my name is Michael Wu. I'm the co-founder and CEO of Amber Group. Uh today my sharing topic is financial services in the agent economy and how it will evolve in my opinion from centering around platforms to centering around autonomous agent native operating systems. Uh but before I start uh I know given the date today every big claim sounds a little bit suspicious.

So let's stick to what's real uh what's life what's useful uh what's already happening uh in financial services what's becoming technologically possible and what kind of companies will matter in this new era. So my view is that we are entering an AI agent-driven economy that will very soon surpass the Lexi economy. Uh and financial services or finance will have to look very different in this new agent economy.

For the last decade or so, finance or fintex mostly competed on the interface layer, meaning they competed on better front end, better UX, better user acquisition channels, better on boarding experience, all kinds of better dashboards. But in the next decade, that will change. The real advantage will move below the interface layer onto the operating layer um to what I think will be agent native operating systems that coordinate from user intents to product selection to liquidity to execution to

controls and basically the entire service workflow end to end. In other words, finance is going to move from standalone platforms towards shared agent native operating systems. And the analogy I often find the most useful for my audience is AWS. So if you think about it before the likes of AWS and other cloud servers what when you had to run a website it was to run a local server you had to configure the servers yourself run all the workloads locally and all of that and essentially

uh the likes of AWS and the cloud services they abstracted that away. So post AWS essentially now running a web page or running a web app is just build the app or build the website and run on cloud. So that abstracted away not only the complexity of running web servers, but economically they just coordinated resources much more efficiently and usually cloud servers are more reliable than your local servers in terms of uptime in terms of all the security guard rails. They're better and they make starting a web-based app or a web business a lot easier. I actually think

without the likes of cloud servers or AWS, I don't think we get to the scale of internet economies today and I think something similar is going to happen in finance. My thesis is that essentially in this agentdriven economy, financial services will also run on these shared agent native operating stack just like how web apps or web businesses run on shared cloud services today. And crypto digital assets, crypto, whatever you call it, is naturally the economic rails embedded in these agent native operating systems.

So as a founder CEO of Amber uh my ambition is never to apply AI onto yesterday's products. I want to build the financial stack for what comes next. So why is this shift happening now? The shift is happening now because three curves are intersecting. First, AI has moved from chat bots to workflow engines.

Two or three years ago, most people still interacted with the AI as a chatbot, right? Like uh like the name chat GPT suggested literally. But now it's all about AI agents and workflows. I mean like many of us here today probably myself included. Yesterday we all got our own locally replicated cloud code, didn't we? So things are moving this way. AI has already gone from just answering questions to actually doing end to end work. Specifically in finance, this means automated workflows in a lot of companies like ours are already reviewing on boarding docs,

routing client requests, generating reports, risk memos, facilitating executions. And secondly, crypto or digital assets have matured from a narrative to now finally an infrastructure. I think for a long time our industry sold crypto as a narrative or a thesis, a macro thesis, uh a generational wealth thesis, sometimes a cult thesis. Uh but I think what finally matters now is more about the actual digital asset infrastructure because now we are finally I think since the last two or

three years start to see digital asset rails being used for more mainstream financial functions such as standardizing different assets and collaterals onto the same ledger instant global transfers 247 global settlements stable coins, tokenized exposures for different assets and so on. It's a big upgrade honestly in my opinion that this is much better than the traditional financial real in terms of the traditional financial real was built with or around a lot of human constraints.

All the limit market hours because people need to get off work. All the delayed settlements because there are routes of settlement process remain uh required human interventions. Uh fragmented ledgers because different regulators different jurisdictions want to keep them separate. Trapped collaterals sometimes intentionally sometimes technologically set up that way and all the reconciliation work that are inefficient ineffective expensive full of mistakes that happen due to all of that.

So it's very natural to me that for AI agents and upcoming agent economy with all their autonomous workflows, crypto is naturally the much better rail for this economy. And third, I think financial institutions are also adapting and therefore they now demanding two things at once, speed and trust. Financial institutions now want more automation but not at the cost of losing controls. Financial institutions now want more intelligence but not at the cost of auditability.

Financial institutions want more speed but not at the cost of governance. So they now require both. It's no longer enough to just be fast but opaque. It's not no longer enough to just be compliant but very slow. It's no longer enough to be intelligent but uncontrollable unuditable ungovernable. That will be the new operating standard in finance. Financial systems or financial players will be required to be highly automated, clearly governed and auditable.

So to make all of this more concrete, I use this model called the four layers of finance. And I will go through each layer one by one and share with you what I think they will look like in this agent economy. The first layer is distribution. Distribution is channels, licenses, branded platforms. These are the Binances, Coinbases, upsets of the world. These are the license holders of Mika, VASP or whatever you call it in each country. I mean we we we have those as as well within Amber um for example in Singapore in Dubai and all of that.

So historically the distribution layer has always been about funnel. They about user acquisition funnels. They about acquiring users, converting users, onboarding users, supporting users and the distribution layer. The platforms will continue to do so. But I think increasingly they will have even simpler, easier and smarter conversion and monetization once they have the users because of these shared operating structure that will that will do the remaining work.

And that leads us to the second layer, the services layer. This is where the autonomous agent native operating systems actually come to play. services layer is where the actual financial services happen. This is if you run a market making business the liquidity provision mandates actually gets matched. This is if you run asset management the products actually got selected and matched and riskmanaged. This is if you run a brokerage business the products actually or the the brokerage orchestration actually happened.

And again, you know, I think this is where a lot of fintexs are still mistakenly thinking in terms of screens. They're still thinking about a feature, adding a feature, some UI tweak, a mobile app, click this button, and that happens. But let me tell you, AI will very soon replace most of these UIs anyway. And the code is cheap nowadays. So the marginal cost of personalization for every user will very quickly trend towards zero.

So the competition will become can you actually seamlessly convert end to end from user intent to product selection to pricing to execution to post trade settlements to reporting all in one coherent workflow and do that reliably repeatably and auditably. This again is about operating system. It's not about the UI. And a truly good agent native operating core will naturally be able to offer much more personalized and optimized financial services to every user than any standalone platform or app can ever offer. This is where the new economy of

scale happens. This is where the edge will compound for successful players in the new era. The next layer is the risk-taking layer, the trading layer because I mean this is also where you know the DENS market makers coming in in crypto usually everyone's a bit of both. So in terms of principal risk taking from actual trading decisions hedging and bearing the NPN from hedging position management collateral management different kind of risk management AI here can improve the speed the research the execution but the risk ownership still sits with the human

because whether you trade manually by click your mouse or you trade through an agent it's going to still going to be you the human or the humans behind that bear the end P&L. So the risk ownership doesn't shift away with technology. The final layer is governance. This is where the compliance, legal control, policy enforcement, approvals, audit, reporting and all of that come in. But this will be quite different as well in this new era because I think today a lot of fintexs especially in crypto I would say still think about governance as a growth constraint. But I think in this

new world, governance will become almost like a prerequisite for growth. Because if you think about it, in this agent-driven economy, in this agent-driven cycle of finance, trust itself is going to be in the architecture. What do I mean by that? If your financial services is not inherently auditable, is not traceable, is not reliable, is not governable, they cannot be trusted by the agents. the agents will naturally have guard rails against accessing financial services like that and if you can't be trusted by the agents you can't be trusted by the humans or the money actually behind

those intent so the these are the these are my four layers of finance distribution services risk takingaking governance and I do believe um agents will transform all four layers more or less and if my model is at least directionally correct then I think at least two conclusions follow. First again agent native operating systems will have a vast advantage and they will beat standalone products. They will beat standalone platforms because today's platforms or at least each product each function on the platform is trying to solve one specific task but an agent native operating system

does it from end to end. It's actually trying to coordinate tasks and converts user intent highly personalized user intent directly into complete results from these intent. Secondly, actually counterintuitively especially for our industry, I think the control panels and guardrails that slow down the services that slow down the intelligence will be almost as important as the intelligence itself. Because if you go through institutional finance, you'll find most of the time the most important question is often not how smart is the system we are adopting.

It's rather how can the risk be ob observed or traced? Does the system have audit trails can be governed by certain policies or sometimes even simply who can stop this? Who can stop this thing and who takes the accountability if this thing fails. So I do think going forward in finance especially in highly automated finance the best operating systems need to be both efficient intelligent without saying and they need to be highly auditable and governable and uh this is a sneak peek of how we are building the financial stack for the

agent economy at Amber. First again we have our own distribution. We have our own license platform called Ember Premium. We use Ember Premium both as our distribution layer and very importantly as a proof case for our agent native operating systems which are the auese at the bottom. Um our Aues are actually system we use to run our businesses at scale. We use a system called uh AMM for our market making business. We use a system called a a fund for our asset management business.

We use a system called a flow for our client trading business. And we try to build or at least iterate these operating systems as agent native from day one because we build our agents that automate the workflows to operate them together. Again, today these are not fully polished. I don't claim that. But imagine when we are ready to open them up. Then in our vision it will become you anyone as long as you have your distribution you have your users your clients you can run the same amberlike financial services business whether it's asset management brokerage market making whatever as long as you

have clients you have these flows you have these client intents you can use our shared infrastructure the shared operating stack to monetize them just like web apps run on cloud servers So to recap, I think we're moving from an interface economy to a workflow economy in finance. In the interface economy, again, because engineering used to be expensive, personalization was often impossible. So the old economy of scale comes from better interfaces that meet the largest overlapping user demand.

So winning in the interface economy was about better front end, better UX, essentially better surface level convenience for users or better top of funnel client acquisition usually due to marketing due to branding or due to license. I think in the workflow economy personalization will become the default option because again engineering will be cheap and the leap from ideas to products became almost instant. So users will interact with financial services through agents and expect their intents to be converted into results end to end.

In this kind of workflow economy, I think winning comes from actual better end-to-end workflow orchestration and the better policy aware automation and better control, better auditability and therefore better trustability. So with AI today really code is cheap and the leap from ideas to products is almost instant. Today any team almost any team can vive code and demo you some kind of AI agents for finance or for trading or financial automation. That is no longer the hard part. The harder part is whether you can show these workflows, these automated workflows that actually

touches money, that actually has a large financial impact are battle tested. They're institutional scale. They're repeatable. They're controllable. They're governable. especially under market stress and regulatory scrutiny. I think that would be a very different bar and in my opinion it's going to be a much healthier bar because finance really I mean if you want to be a successful fintech going forward it will no longer be about like is your app good. The question should become is your system robust enough to coordinate user intent, the financial complexity, the risk

management at scale, reliably, transferably and accountably. So this is what great fintech should look like and it will in this new era. So what what does it mean for builders, for institutions, for policy makers? So if you're a builder, I do think it's time to stop optimizing for interfaces. It's time to put more energy into design better workflows that solve the entire problem reliably at least in finance.

And if you are in financial institution, I do think you will start to evaluate these automation, these workflows on their policy reliability, on their resistance, on the stress. And if you're a policy maker, this is not the moment to be even more defensive. This is the moment. This is your chance to design standards that can allow innovation, but also leave room to actually hold innovators accountable because this takeover of the aging economy will happen anyway. We will all be better off positioning ourselves for this inevitable future.

So, let me close with this. Digital assets will become this new rail. Agents will automate most workflows and the shared agent native operating system will be the core infrastructure. Uh Amber, this is our ambition. We want to build the financial stack for this upcoming agent economy and I do enjoy every step on the journey towards that. Um thank you and we'll take any questions if we still have some time. Yeah, we ran out of time to question but thank you for the talk. Really interesting topic.