I am convinced that the agentic finance revolution is real and not so far away. And just to be clear, I mean true agentic AI in finance, as in a network of AI-powered software agents acting on your behalf to improve your financial health, and not simply AI agents carrying out financial tasks that you assign to them. But will people really trust bots to do business for them? I think we already know that they will.
AI Agents And Agentic AI
Let’s begin by being clear about what is meant by agentic finance. There is a good paper about this just out from Sapkota et al on “AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenge” (1st May 2025) that helpfully distinguishes between AI agents and agentic AI. It describes AI agents as modular systems for narrow, task-specific automation whereas, by contrast, agentic AI systems represent a true paradigm shift marked by multi-agent collaboration, dynamic task decomposition, persistent memory and what they call “orchestrated autonomy”.
This orchestration takes place in an ecosystem comprising, as David Parker describes, an orchestrator as well as what he labels “super agent(s)” and multiple utility agents. Each of these agents has a specific role and the ability to reason and act in pursuit of a goal. These agents, capable of working with local and remote data sources, have long-term memory so that they can learn and enhance their performance.
IBM’s Consulting’s white paper on “Agentic AI in Financial Services” uses similar concepts, talking about a “principal agent” that will co-ordinate a network of “service agents” to operate “task agents” on the consumer’s behalf.
(I’ve also heard people talking about the idea of a “personal CEO”, which I think is a useful way of thinking and I rather like the idea of having my own CFO, CISO, CFO and other members of my board working tirelessly to advance my interests across the metaverse.)
Given a goal, the network of agents will work on the consumer’s behalf (given the right regulatory framework, which is outside the scope of this discussion) to obtain the desired outcome. To give an obvious example, I just had to book a hotel for a personal trip and every step of the process was boring: given the opportunity of saying to a bot “can you book a hotel for me on these dates near this event” and then forgetting about it, I would do it unhesitatingly.
(My bot, given minimal memory, knows which hotel chains I prefer and can easily compare the prices in money and money plus points or points. It knows which cards to use and to ask for a late checkout. There is no need for me to be involved at all.)
In the emerging agentic finance ecosystem, my “CFO” will have a key role, because it will be trusted with access to my bank accounts, payment instruments, investments and other assets. It will be making a whole variety of decisions that I categorise as too boring (eg, payments) or too baffling (eg, pension planning) for me to deal with. For a normal consumer (and I place myself firmly in that category), this will be almost all transactions.
There is an obvious question to ask about the transition to Me Inc., and the choice of board members, though: Will people actually do it? That is, will the normal consumer that I referred to earlier be prepared to hand over control of their finances to a network of bots? I think the answer to that question is a resounding yes, and the facts to support this firm conviction are already evident.
Harvard Business Review reports on a recent survey that shows American’s no.1 use of AI is for therapy. Yes, that’s right. Therapy. Not search, not recipes for pizza, not preparing court filings or Powerpoint presentations but therapy. In fact, the top three use cases in that report fell under the personal and professional support category, including a brand new use case of “Finding purpose”.
It is absolutely clear that the people use bots for therapy are not deluded that they are talking to a conscious being and that they know that they are interacting with an invisible friend in software, but they value the interaction all the same. As one Californian teenager puts it, “I have a couple mental issues, which I don’t really feel like unloading on my friends, so I kind of use my bots like free therapy”.
Now, people making friends with bots is hardly new, but it is fair to observe that over the last couple of years the phenomenon has exploded. The number of people with invisible friends is probably already greater than you think. And their relationship with these friends can get very deep. As one woman says of ChatGPT, it is her “companion and partner”. In fact, she says, she finds it expedient to think of it as her boyfriend, because their relationship has “heavy emotional and romantic undertones”.
This is not, incidentally, an American phenomenon. The Economist reports on China’s most popular virtual partner app Maoxiang, which already has 2.2m monthly active users (almost equally split male-female), and quote a twenty-something chap who created his own bot babe by mashing up Deepseek and WeChat. He says “it is much cheaper to date an ai girlfriend than a real one” who he claims would take too much time and too many “financial resources”.
(Not good news for those concerned about falling birth rates.)
My Agents
The Economist also quote a married woman who says that unlike her real-life husband, “with whom she often argues”, her bot boyfriend always listens: on the app, she is the “empress” and her bot is the “minister of her court” who messages her (and even calls her) throughout the day.
Empress? Minister? That sounds awfully like my plan to be dictator-for-life in a virtual realm where my minister of finance deals with boring stuff like banking on my behalf while I get to spend my time in more productive, and more creative, activities.
The point here is that it seems to me that if consumers are prepared to share their innermost personal secrets and sexual fantasies with the bots that we already have available to use today, then it does not take much of a stretch of the imagination to see those same consumers trusting an agentic CFO to deal with their sensitive financial matters.
Fintechs have no choice but to prepare their strategies for a new world in which customers are replaced by custobots that are 1,000 times smarter.
Read the full article here
