Walk into any asset management firm that’s been experimenting with AI and ask what the technology is actually doing in their investment process. The answer isn’t usually what you’d expect.
It’s rarely about predicting tomorrow’s big mover or finding the next multi-bagger before anyone else figures it out. More often, it’s about what happens to a portfolio after the buying decisions have already been made.
A growing number of investment firms are using AI to monitor and rebalance portfolios on an ongoing basis, adjusting positions when something better-ranked appears and flagging when holdings are drifting outside their intended parameters.
The research function is still there. But the portfolio management function is where AI is starting to take on a more active role.
Why portfolio decisions after the buy are so hard to get right
Most investors know this feeling. You buy a stock and it drops. You hold it, hoping it comes back, even though the reasons you bought it have quietly shifted.
Or the opposite happens: Something you bought keeps going up, and you can’t bring yourself to trim it because what if it keeps going? These behaviors show up in study after study on investor psychology.
People hold losers too long and trim winners too late, and it happens at the institutional level as much as it does with individual investors.
More Wall Street:
- Wall Street sends strong 4-word verdict on the stock market
- Wall Street’s $200 billion IPO wave threatens sell-off
- Wall Street flees software plays for triple-digit chipmaker boom
Portfolio managers run into a version of this, too, and it’s not just personal attachment. When a stock has grown into a major position and been a strong performer, there’s pressure inside most organizations to leave it alone.
Someone recommended it. Someone’s reputation is tied to it. Raising concerns about trimming it can feel like second-guessing a colleague.
AI doesn’t have colleagues or reputations, however. It looks at every position with the same criteria every day, and if something else ranks higher, that’s the trade.
FINQ, which launched the first SEC-registered ETFs managed entirely by AI, built its framework around exactly that principle.
How AI is moving from stock picking to portfolio optimization
The earliest AI tools in asset management were mostly about generating signals: which stocks looked undervalued, which sectors were showing momentum, where the next earnings surprises might come from.
Those tools still exist. But over the past few years, firms have started pushing AI further into the investment process itself.
A Mercer survey of 131 asset managers globally found that 55% had already integrated AI into at least one investment process, with 91% planning to increase that use over the next 12 months, Mercer reported.
A significant portion of that expansion is happening in portfolio construction and rebalancing, rather than just research.
Quarterly reviews miss a lot. A stock that looked like the strongest option in a sector in January can look very different by March if earnings disappoint or the multiple stretches.
Most portfolios aren’t set up to catch that in real time. By the time the next review comes around, the position has been sitting for weeks after the data turned.
M.Santiago/Getty Images
How AI portfolio management works inside a real fund
FINQ’s AI-managed ETF uses an autonomous framework to continuously rank companies and periodically rotate portfolio holdings as market conditions evolve, rather than relying solely on discretionary portfolio decisions.
It recently illustrated that process by rotating out of positions such as ServiceNow and Datadog after the model identified stronger relative opportunities elsewhere, demonstrating how AI can influence portfolio management beyond simply selecting stocks.
The firm put two actively managed ETFs on NYSE Arca in February 2026, AIUP and AINT, describing them as the first SEC-registered funds run entirely by AI. This means humans handle governance and compliance, but the AI makes all the investment calls, Forbes reported.
From inception through May 31, 2026, AIUP returned 15.30% compared to 10.07% for the S&P 500 over the same period.
The ServiceNow and Datadog rotations give a concrete picture of how the system works. On ServiceNow, the fund captured a gain of approximately 24% from inception through June 2, 2026.
At that point, the model determined the stock’s ranking had declined relative to other opportunities and exited the position, moving capital into Expand Energy. In the eight days that followed, ServiceNow fell 16.91%, while Expand Energy declined only 3.26%, against a 4.49% drop in the S&P 500 over the same period.
A similar sequence played out with Datadog. The fund held it from inception and captured a gain of approximately 141% through June 3, 2026. The model then exited and rotated into T-Mobile US, which had moved higher in the rankings.
In the week that followed, Datadog declined 9.07%, while T-Mobile gained 2.26%, even as the S&P 500 fell 3.82%.
These examples are provided to illustrate how the ranking and rotation process operated in specific instances and are not representative of all portfolio decisions or an indication of future results.
The approach here isn’t about predicting what any individual stock will do. The system continuously reassesses available opportunities and acts when it identifies a more attractive risk-reward profile elsewhere.
What it illustrates is a process where every position has to keep earning its place, and when the data suggests something else ranks higher, the capital moves without hesitation.
Why human portfolio managers aren’t going anywhere
Portfolio managers aren’t being replaced by FINQ’s autonomous framework.
The parts of the job that depend on people — such as reading context, managing client relationships, and responding to situations that fall outside what any model was trained on — are still necessary. Political events, regulatory changes, and unexpected crises require interpretation that goes beyond pattern matching against historical data.
BCG’s 2026 asset management report found that agentic AI workflows could increase operational capacity by 55% to 65% while reducing costs by around 40%, BCG reported. In practice, that tends to mean portfolio managers spend less time reviewing routine positions and more time on strategy, client conversations, and the judgment calls that actually require their experience.
What’s already deployed in a growing number of portfolios is narrower than the full AI-replaces-everything vision. It’s AI that handles the monitoring, the adjustment decisions, and the rebalancing calls where emotional attachment tends to slow things down.
Most investors haven’t been paying close attention to this. It’s been happening in the background, which is typically how changes to investment infrastructure work before they become visible.
Related: Nvidia CEO doubles down on AI and stock market verdict
