Agent AI and the economy “do it for me”

Davos this year – more appropriately, the annual meeting of the 54th World Economic Forum (WEF) in Davos, Switzerland, was impossible to escape AI. All conversations about hydrogen production, loyalty schemes, horse training and ethical investment touched on AI. The central focus of these debates was the shift towards “agent AI” that has real meaning in the world of fintech.

Agent AI in finance

New report from Citi on Agent AI – Finance and “Do it for my economy (disclosure: I’m one of the external experts cited in the report) to the demands of future generations of AI customers The financial services sector that needs to be addressed.

The shift towards agent AI is clear. Speaking at the Salesforce Annual Conference last year, CEO of Marc Benioff said that AI agents are the next step in the evolution of AI. He is certainly right, and as financial services platforms move from providing embedded finance through APIs to embedded financial services via agents, they will be able to reach the services offered and the business models behind them. There is a true revolution. One of the key fintech investors, Bain Capital Venture, is equally bullish. They are just as destructive as the advent of the e-commerce and M-commerce era, so the agent shopping era is just as destructive, raising people on the front-end of these changes, while others who are lagging behind. I’ll take it down. Financial organizations that rely on the business models I grew up in, such as exchange fees and net interest margins, need to rethink their strategies.

Therefore, Agent AI is clearly a key trend in 2025, and this is not just the views of engineers. Citi’s report reveals that agent AI is where smart money is moving forward, with the exception of sector-wide hype. Last year, more than a third of VC funds were sent to AI startups, with autonomous agents and digital colleagues achieving the greatest growth in VC trading activities (continued by Genai for customer support operations). They also say they expect to see continued growth in investments around Agent AI throughout the year, based on recent encouragement from BigTech leaders, startup entrepreneurs and the VC community. So it’s easy to see why the AI ​​agent market is currently projected to grow from around 5 billion to nearly 50 billion by 2030.

(VC is recalibrating the entire investment paper on Agent AI. The software business pie is growing dramatically, and smart money is a way for Agent AI to create more unicorns than mobile or cloud computing. I’m betting.)

Progress is rapid. Shortly after Davos, Openai launched an “Operator.” It is an AI agent that allows you to control your web browser to automate tasks such as travel accommodation booking, restaurant booking, online shopping, and more. The operator is equipped with a Computer Use Agent (CUA) model that combines the vision capabilities of OpenAI’s GPT-4O model with the inference capabilities of the more advanced model. What this means is that agents do not need an API to access the service. You can use the buttons to navigate through the menu (probably you can pass in the “You’re not a robot” puzzle).

With the advent of CUA, we can see the practical evolution of full-scale agent commerce in a strategic time frame. Agents simply get the service online, look for the agent API, then if no such API is found, they simply go to the web page as the human customer does. From there, there is a small step to an agent that gets services from other agents rather than through a “traditional” web interface or API! The agent pays to other agents and bypasses the existing payment path completely. True “A-commerce” is here!

However, as I anticipate, operators need human supervision of certain categories of tasks, including financial transactions. However, this is a temporary step only needed because there is a lack of the digital identity infrastructure needed to enable fully autonomous transactions.

Patricia Russo has set out clearly in his excellent work on LinkedIn, and ultimately means an audit trail for an agent to end with his digital identity, develop his reputation, and of course, when things go wrong It becomes a matter of responsibility. This means that the agent must have its own credentials, rather than “borrowing” the consumer’s credentials that the agent represents. Giving the agent authority to act on your behalf in certain circumstances, like Russo points out, adding authorized users to your credit card. Agent providers are responsible for the agents’ actions, but by assigning identity and providing credentials in this way, it makes it easier to engage in commercial transactions with manageable risks and thus promote competition.

The transaction proceeds based on approval. For example, if an agent appears at my bank and asks me to transfer money, the bank needs to identify the agent and discover whether the agent is authorized to carry out a particular transaction. This can be done by having the agent’s database and its permissions, but a much better way to do that is that the agent must present the relevant credentials. That credential includes the agent’s public key and is signed with the bank’s private key within the framework used. My bank knows the public keys of other banks, so it’s easy to check your credentials. There are already some really interesting ideas in this field (for example, the mental framework).

Agent means opportunity

Big Tech is planning to provide its customers to its own agents, and it is inevitable that many customers (for example, me) cannot hand over so many decisions about financial transactions to the agents. Agents acting on behalf of me, the ultimate customer of financial institutions (and for my greatest benefit) will become consumers of financial services for those institutions. What do these AI consumers want? I will continue to go back to this question. Partly because I don’t know what the answer is, I think it’s a field of strategic importance (and existential threats) for financial institutions, and a field of potential opportunities for fintechs. is.

(And frankly, it would be fun to see the growth of non-human customers change the way we do our online business.)

When I stopped by the dome of MIT in Davos, I sat on a very interesting panel about international AI development. This is a panel that includes Deepmind’s Nobel Prize winner Ir Demis Hassavis. Ir Demis said he was “cautiously optimistic” about the future and that he made me feel the same way. In the end, I came back from Davos. Two things, wearing a wool-like hat is a real excitement about the digital finance opportunities in the upcoming world of A-commerce.