I am writing this column sitting at a café. In front of me, two students compare two ETFs on ChatGPT and debate the virtues of one or the other. AI is everywhere in their actions: in the way they seek information, in the funds they look at, in the underlying assets in which they plan to invest.
They did not open the regulatory documentation of the funds. A priori, they have not seen either a private banker or a financial advisor. They interact in threes, with an AI which serves as a search engine, analysis assistant and implicit trusted third party.
We’ve seen this movie before. At the end of the 90s, the Internet gradually entered the economy. The prophets announce the death of physical commerce, the advent of direct democracy, the end of intermediaries. Skeptics sneer at what they believe to be a passing fad. Twenty-five years later? The Internet was neither the apocalypse nor the messiah. Just a great accelerator.
Generative AI is following exactly this trajectory. It amplifies what already exists, both the strengths and the flaws. When Morgan Stanley deploys a generative AI model for its 16,000 advisors, it is not reinventing finance. It turbocharges the existing. This nuance changes everything.
AI as a magnifying mirror
Think of AI as a mirror that amplifies without distorting. It does not create magical knowledge. It reorganizes what is already lying around in the digital ocean: buried analyst reports, unreadable tax case law, scattered market histories. Synthesizing in three seconds what a human would take three days to compile, such a reduction in time (and therefore costs) is not without consequences.
For an advisor overwhelmed by regulatory complexity, MiFID II, GDPR, Loi Pacte and their cohorts of decrees, it is a cognitive aid. McKinsey estimates the time saved on certain administrative tasks at 30-50%. Finding a 2019 tax clause, comparing twenty euro funds, cross-referencing ten years of macro data: dedicated tools transform these chores into trivialities.
For the saver, the promise is tangible. Done, meetings in two weeks for a simple question. Finished, the approximations on the optimal tax envelope. AI brings fluidity where there used to be bureaucratic friction. Available 24/7, documented, responsive.
But there you go. This fluidity has its downside. By speeding up everything, AI can short-circuit thinking time. By answering everything, she risks stifling the only question that matters: “Is this really what I want?”
The score without the performer
A perfect score is beautiful on paper. All the notes are in tune, the tempo is exact, the harmonies are mathematically optimal. However, a score is not music. It takes an interpreter to know when to speed up, when to slow down, when to respect rigor or take creative liberty.
The “augmented advisor” is exactly that. The expression is correct provided the roles are not reversed. AI must remain the tool. The human, the pilot. The temptation is great to do the opposite.
A study from the National Bureau of Economic Research shows something fascinating: AI raises the quality of junior advisors’ answers (real progress) but it standardizes everyone. The mediocre can improve. The excellent ones can lose their uniqueness. The variance collapses.
That’s the trap. AI models are probabilistic machines. They are looking for the most statistically consensual answer. Not necessarily what’s optimal for you, but what’s optimal on average for someone like you. AI excels at producing basic advice: diversification, investment horizon, risk-profile matching. The solid technical base. The bare minimum of good management.
But the nuances? The desire to transmit, the need for psychological liquidity, the bet of conviction on an emerging sector, these details which make the difference in an asset? The algorithm misses the point.
A compass does not hold the bar
Let’s return to the individual level. AI is the ultra-advanced GPS which calculates the optimal route based on weather, currents and fuel consumption. Tremendous. But it is a navigation tool, not governance.
The captain decides whether to really take this route or go around the storm even if it takes longer. He knows that his crew is tired and that they need to make a stopover. He can choose to deviate to help, even if it is not optimal on paper.
For savings, it’s the same. AI can optimize your asset allocation on a technical level. But she can’t decide for you what really matters. It calculates that investing in the stock market maximizes your expected earnings over twenty years. But she doesn’t know that you sleep poorly with volatility. She tells you it’s fiscally stupid to withdraw money now. But it ignores that you have a vital need for a project which has no economic sense but all its existential sense.
This is where the real revolution lies, if there is a revolution. AI forces roles to be clarified.
An accelerator, not a replacement
So, buzz or revolution? A bit of both, and mostly neither.
Yes, AI is tangible progress. It democratizes access to quality information, reduces processing costs and streamlines the experience. For someone who simply wants to open a PEA and understand the basics of diversification, it is a valuable ally. It breaks down the barriers to entry.
Yes, it raises the average level of advice by eliminating gross errors (bad tax calculations, profile inconsistencies) and enriching junior advisors with instant access to collective expertise.
But it has structural limits. She extrapolates the past more than she anticipates ruptures. It favors consensus over originality. She feigns empathy without understanding emotion. And above all, it poses a question of governance: who controls the algorithm? Who benefits from optimization?
Understand rather than delegate
The saver and advisors of tomorrow will have to master a new art: asking the right questions to AI, challenging its answers, distinguishing technical optimization (where the machine excels) from the life project (where only humans decide). It’s less comfortable than delegating everything, but it’s infinitely more powerful.
Human advice is not going to disappear. He will recompose himself. Its value will no longer be in the restitution of information. It will be in three irreplaceable dimensions: critical judgment, knowing how to say no to the machine; emotional intelligence, supporting moments of doubt; and heritage creativity, imagining non-standard solutions for complex situations.
AI is neither the messiah who will democratize excellence, nor the charlatan who will ruin savings. It’s a revealer. It reveals the weaknesses of the industry but also its potential: accessibility, efficiency. It reveals the saver’s fragilities but also their new room for maneuver: cognitive autonomy, increased demands.
This lucidity will make the difference. Between those who will experience AI as an inevitability and those who will use it as a lever.
The students in front of me, by having a three-way dialogue with the AI rather than accepting everything that is written, have basically already understood several things: the AI is already there, we might as well deal with it.
The question is not: “Will AI change savings?” » She is already changing it.
The real question is: how are we going to use it?
