This is the third in a three-part series. Start from the beginning: Where does proprietary value live in a frontier AI world? or read part two: The case for the data layer: the part of AI most firms forget to build
There is a version of AI adoption in investment management that looks impressive and delivers very little. A firm connects its CRM. It subscribes to a data provider. It runs a few prompts, gets some outputs, and calls it a pilot. The data is all there. The model is capable. And yet nothing in how the firm operates has materially changed.
The problem is not the data or the model. It is the absence of anything connecting them to a decision.
The orchestration layer is where institutional knowledge about how a firm actually operates gets encoded into the system: the investment thesis that shapes which signals matter, the workflow that routes a flagged company to the right analyst, the process that ensures a meeting briefing is ready before the calendar invite fires. Without orchestration, data sits. With it, data moves.
This becomes clearer when you consider where sophisticated deal teams actually spend their time. It is not finding data — most firms have access to more than they can usefully process. The real cost is in synthesis: pulling together what is known about a company across internal and external sources, connecting it to current context, and turning it into a position someone can act on. That process, repeated hundreds of times across a portfolio and a pipeline, is almost entirely a problem of orchestration, not data availability.
A CRM that captures every interaction a firm has with a company is genuinely valuable. But the value is not in the capture. It is in knowing when a change in that record is material, surfacing it to the right person at the right moment, and prompting the right next action. The data layer provides the raw material. The orchestration layer makes it useful.
Every time a deal team uses an orchestrated workflow, something is learned. The system develops a more accurate picture of which recommendations the firm acts on, which it ignores, which sectors it is actually interested in versus which it says it is. Over time, the orchestration layer adapts to the idiosyncrasies of how a specific firm operates — something no generic tool can replicate. That adaptation is not a feature. It is a moat.
In a market where every firm has access to the same frontier models and the same data providers, the differentiator is not what you can access. It is what you can do with it, how quickly you can act, and how much institutional knowledge shapes the quality of that action. Two firms can have identical data subscriptions and identical AI tooling and produce radically different outcomes, depending on whether the layer connecting them to their actual work has been thoughtfully built.
The orchestration layer is where domain expertise lives: the workflows, agents, and task structures that encode how investment professionals actually work. That knowledge cannot be bought from a vendor. It has to be built, over time, in close proximity to the teams doing the work.
Data is what a firm knows. Orchestration is what a firm does with what it knows. The firms that figure out the second part will be the ones that look back in five years and understand why the gap opened up.
If you're thinking about how your firm's data and workflow infrastructure needs to evolve, we'd be happy to talk through what that looks like in practice. Book a demo
Read the full series from the beginning: Where does proprietary value live in a frontier AI world?