Agentic AI And The “Do It For Me” Economy

At this year’s Davos — or more properly the 54th Annual Meeting of the World Economic Forum in Davos, Switzerland — it was impossible to escape AI. Every conversation — whether it was about hydrogen production or loyalty schemes, horse training or ethical investment — touched on AI. Central to these discussions was the shift toward “agentic AI,” which has real implications for the world of fintech.

Agentic AI In Finance

A new report from Citi titled “Agentic AI: Finance & the Do It For Me Economy” talks about how agentic AI that adapts to circumstances and acts autonomously has huge implications for a financial services sector that will need to respond to the demands of the coming generation AI customers. (Disclosure: I am one of the external experts quoted in the report.)

The shift towards agentic AI is clear. Speaking at the Salesforce annual conference last year, CEO Marc Benioff said that AI agents are the next step in the evolution of AI. He was surely correct, and as financial services platforms shift from providing embedded finance through APIs to embedded financial services through agents there will be a genuine revolution in the services offered and the business models behind them.

One of the key fintech investors, Bain Capital Ventures, is similarly bullish. They say that like the emergence of the e-commerce and m-commerce eras, the agentic shopping era will be similarly disruptive, elevating some who are on the front-end of these changes while demoting others who lag behind. Financial organizations that depend on the business model opiates that I grew up with, including interchange fees and net interest margins, will need to rethink their strategies.

So, then, agentic AI is clearly a key trend for 2025 and this is not only the view of technologists. The Citi report is clear that, aside from hype across the sector, agentic AI is where the smart money has been going. Last year more than a third of VC funding went to AI start ups and autonomous agents and digital co-workers had the biggest growth in VC deal activity (followed by GenAI for customer support operations).

They also say that based on the recent encouragement by BigTech leaders, startup entrepreneurs, and the VC community, they expect to see continued growth in investments around agentic AI throughout this coming year. Hence it is easy to see why the AI agents market is projected to grow from some five billion now to almost fifty billion by 2030. (VCs are recalibrating their entire investment thesis around agentic AI. The pie for software businesses has just grown dramatically, and smart money is betting that agentic AI will create more unicorns than mobile or cloud computing did.)

Progress is rapid. Just after Davos, OpenAI launched “Operator” which is an AI agent that can take control of a web browser to automate tasks such as booking travel accommodations, making restaurant reservations and shopping online. Operator is powered by a computer-using agent model that combines the vision capabilities of the OpenAI’s GPT-4o model with reasoning abilities from their more advanced models. What this means is that an agent doesn’t need an API to access services, it can use buttons and navigate menus (and presumably pass the “prove you are not a robot” puzzles) just as people do.

With the advent of the computer-using agent, we can now see the practical evolution of full-blown agentic commerce in strategic timeframes. Agents will go online to obtain services, look for an agentic API and then, if no such API is found, simply access the web pages as a human customer does. From there it is a small step to agents obtaining services from other agents, not via “traditional” web interfaces or APIs at all. Agents will be paying other agents, bypassing our existing payment paths completely: True “a-commerce” is coming!

As it stands though, Operator requires human supervision for certain categories of task including, as I am sure you would expect, financial transactions, so that consumers currently need to take control to enter payment information. But this is an interim step that is only needed because we lack the digital identity infrastructure needed to allow fully autonomous transactions.

As Priscilla Russo set out clearly in an excellent post on LinkedIn, ultimately it will be the issue of liability that mean that agents will end up with their own digital identities, to develop their own reputations and of course audit trails for when things go wrong. This in turn means that the agents will need to have their own credentials rather than “borrow” the credentials of the consumers that they represent. Giving an agent authority to act on your behalf in certain circumstances is, as Russo notes, rather like adding an authorized user to a credit card. While the agent provider would be responsible for the agent’s actions, the allocation of identity and the provision of credentials this way makes it straightforward to engage in commercial transactions with manageable risks and therefore encourage competition.

Transactions will proceed on the basis of authorization. When an agent shows up at my bank, for instance, and asks to transfer money then the bank needs to identify the agent and then discover whether the agent is authorized to execute the specific transaction. This could be done by having a database of agents and their permissions, but a much better way to do it is require the agent to present the relevant credential. That credential will contain the public key of the agent, signed with the private key of a bank within whatever framework is being used. Since my bank knows the public key of the other bank it can easily check the credential is authentic. There is already some really interesting thinking going on in this space (an example is the Ethos framework).

Agents Mean Opportunities

Big Tech, then, is going to give customers their own agents and it is inevitable that many customers will be only too happy to hand over a great many decisions around financial transactions over to agents, meaning that while I will remain the ultimate customer of the financial institutions, agents acting on my behalf (and, I hope, in my best interest) will be the consumers of those institutions’ financial services.

What will these AI consumers want? I keep returning to this question, partly because I do not know what the answer is and partly because I think it is an area of strategic importance (and existential threat) for financial institutions as well as a area of potential opportunity for fintechs. (And, frankly, it is going to be fun seeing the growth of non-human customers change the way the we do business online.)

When I stopped by MIT’s Dome at Davos, I sat in on a very interesting panel about international AI development, a panel that included Demis Hassabis, the Nobel Prize-winning cofounder of DeepMind. Demis said he is “cautiously optimistic” about the future and he left me feeling the same way. In the end I came back from Davos with two things: a rather fetching woolly hat and real excitement about the digital finance opportunities that abound in the coming world of a-commerce.

Editor’s note: An earlier version of this article misnamed Priscilla Russo as Patricia Russo. It has been corrected.