For years, institutional investors have been promised that automation would clean up the operational mess. It would streamline reporting, speed up reviews and give teams the breathing room they have been missing. And to an extent, it has helped. But the first generation of automation has also shown its limits.
Most AI tools were built for scale, not scrutiny. They are fast, but they do not understand the regulated language we deal in. They can summarise, but they cannot show their working. They offer outputs, but not the transparency that fiduciary environments demand.
In the pursuit of efficiency, many organisations have ended up with tools that answer the wrong problem. They process, but they do not explain. They speed up tasks, but they do not strengthen oversight. That is not a sustainable trade-off for an industry built on accountability.
This is why the conversation is beginning to shift. Across asset managers, pension schemes and consultants, we are seeing a move away from automation for its own sake and toward something more meaningful: autonomy. Not machines acting independently, but systems that understand the context they operate in and support humans with clarity, consistency and a clear audit trail.
In other words, tools that help investors take back control of AI rather than surrender it.
Why generic automation is no longer enough
The truth is that most automation tools were designed for environments where the cost of being wrong is low. They were never built for fiduciary settings, where every judgement must be evidenced, every step must be reviewable, and every input must be traceable.
Recognising a pattern is not the same as understanding a regulated conversation. Matching words is not the same as interpreting a stewardship debate, or capturing the nuance of a manager meeting, or understanding why a change in tone matters.
Automation accelerates processes. Autonomy supports judgement. And the industry is beginning to realise the difference.
What autonomy actually looks like
Autonomy is not about letting technology make decisions. It is about giving people systems that remove friction without removing control.
When we talk to institutional investors about autonomy, the themes are always the same.
They want AI that understands the language of regulated finance, not just the vocabulary of the internet. They want outputs that can be reviewed, challenged and signed off. They want a clear line between what went in, what came out and why. And they want all of this in a form that does not add risk as it adds speed.
These are not technical requirements. They are governance ones.
Domain-trained AI: the shift that matters
The most interesting development right now is the rise of domain-trained AI. This is where intelligence systems are trained on the real conversations, decisions and governance logic of a specific field. In regulated finance, this matters.
It is the difference between a generic model trying to infer meaning and a domain-trained system recognising the patterns of a pensions trustee meeting, the cadence of a private-markets due diligence call, or the structure of a stewardship review.
It is also why models trained on authentic industry dialogue behave differently. Codex, the intelligence engine behind SOFI [the AI tool that captures spoken input and transforms it into industry-grade summaries, minutes, or action records], sits in this category. It learns from the conversations that actually define our industry: advice interactions, investment committee debates and years of real governance practice. The result is a tool that does not just summarize but contextualises. Not perfectly and not magically, but in a way that reflects the world it was built for.
That, in itself, is the shift.
Different domains, different futures
What makes the landscape more interesting is that the investment chain is fragmenting into distinct domains of intelligence.
Public markets teams are looking for tools that can track engagement quality, capture meeting signals and create consistent evidence trails. Private markets teams need help with long, qualitative due diligence processes where what is said – and what is not said – matters as much as the numbers. Adviser environments are grappling with Consumer Duty and need structured reasoning they can defend.
These domains do not compete. They reflect the reality that different parts of the investment ecosystem need different types of intelligence. And the future will not be one model that does everything, but multiple specialised systems built around the governance of each space.
SOFI happens to show what this looks like on the adviser side – configurable governance logic, transparent data lineage, outputs that can be challenged and signed off. But the point is not the product. The point is the direction the whole industry is heading.
Why autonomy resonates with institutional investors
The appeal is simple. Autonomy gives teams something automation never could: trust.
Not abstract trust, but practical trust. Trust that the reasoning is visible. Trust that the audit trail is intact. Trust that the system supports governance instead of complicating it.
It gives boards the evidence they need, gives clients the transparency they expect and gives regulators the demonstrable outcomes they ask for. It removes manual pressure without removing accountability.
Autonomy does not replace judgement. It removes the noise that gets in the way of it.
If the industry wants tools it can trust, it has to build systems that reflect the way fiduciary decisions are made. That is the real shift underway.
Autonomy is not a technology trend. It is a governance one. And the investors who embrace it will not just work faster; they will work with more confidence, more consistency and far more control.
