Stop Asking Firms for Data. Start Asking What Intelligence They Get Back

A data-sharing proposition is not a technology decision. It is a value exchange — and in a regulated market the return path has to be designed, not assumed.

We naturally tend to answer the question we are given. But in AI and data strategy, it is often worth first asking whether it is the question that actually needs answering. That distinction can determine whether a shared utility becomes something firms continue to support — or something that gradually loses relevance.

I was recently invited to contribute to the discovery process for a new platform built around firms contributing their own data. The initial discussion naturally included what information would make the proposition useful and how that information might be presented. But the more interesting question sat underneath the platform itself: what would make firms continue contributing to it once the initial novelty had passed?

So before the meeting I wanted to test the question underneath the brief. What decision was the proposition ultimately meant to change? Whose behaviour would it influence? And, perhaps most importantly, would firms still see sufficient value to keep contributing to it three years later? The question became a larger one: if a firm contributes its data, what proprietary intelligence does it get back that helps it run its business better? That is the question worth answering, and it is one this island's AI and data agenda will increasingly need to confront.

This connects directly to where the series began. In Part 1 of Series 1, I argued that AI initiatives should fund decisions, not tools — that the first test of any use case is the real decision it is meant to change. A data-sharing proposition is the same discipline pointed outward: not “what data can we collect?” but “what decision does the resulting intelligence let a contributor make that they could not make before?”

Answer the question behind the question

The instinct that separates useful work from busy work is a simple one: before solving the problem, ask what the problem is a symptom of. Almost every brief arrives pre-narrowed. Someone has already decided what the issue is and reduced it to a request — and that reduction is convenient, not because the requester is careless, but because the person closest to a problem is rarely best placed to see its full shape.

In the case that prompted this essay, the surface discussion was about what to present and how. The question underneath it was about reciprocity. That shift in framing is not cosmetic. It changes what the thing needs to be, who it is for, and whether anyone bothers to sustain it once the novelty fades. And in Jersey it is not hypothetical: the Jersey AI Playbook explicitly names the island's compact scale as a strategic advantage — the ability to align faster, share learning, and potentially build shared infrastructure such as a centralised KYC utility. Every one of those depends on firms contributing something, which means every one of them lives or dies on what they get back.

Reciprocity is the test most propositions fail

When you ask anyone — a person, a team, a firm — to give you something, the honest test is what they get back. Not what you get. What they get. This is the difference between a request and an exchange, and a great deal of institutional work is quietly one-directional: give us your data, your time, your compliance, and trust that it is worth it.

“Free” is the tell. A platform can be free to join and still expensive to feed — in staff time, data preparation, validation, governance, and reconciliation. In a regulated financial-services environment, those activities are not trivial: data quality, appropriate governance, lawful processing and the management of material third-party dependencies all need to be considered within the applicable regulatory and legal framework. A firm weighing participation is therefore not asking “what does it cost?” It is asking “what level of reciprocal value justifies the continuing operational cost of taking part?” Build the return path deliberately, or the whole thing decays the moment the effort of contributing outweighs the benefit of having contributed.

Mind the second-order effects

The most valuable insight is often the most sensitive one — and that is precisely where good thinking earns its keep. It is easy to design the part everyone agrees on. The hard part is the tension the surface brief never mentions.

Two structural problems recur. The first is the cold-start problem: the platform only becomes valuable once enough firms participate, but firms only want to participate once it is already valuable. Solve for that deliberately — by seeding early value — or it never reaches the point where it works. The second is the incentive tension: the intelligence firms would most value (who is winning work, where demand is moving, why they were not considered) is frequently the intelligence the owner is least willing, or least permitted, to return. Noticing a thing is technically possible is easy. Noticing it is commercially, competitively or ethically constrained is the actual work — and in a jurisdiction built on confidentiality and trust, aggregation, anonymity and permission are not afterthoughts.

AI belongs above governed data, not instead of it

There is a strong temptation to skip all of this and let a model narrate the data instead. It should be resisted. The most useful role for AI in a proposition like this is a commentary layer that sits above governed metrics and can always be traced back to them — not an ungrounded narrative invented to look like insight.

The numbers explain why the discipline matters. Gartner reports that 80–95% of organisations have not yet realised financial returns from AI — not because the technology fails, but because the process and the data underneath it never changed. I made the same argument in Series 2 Part 4: AI scaled on weak data foundations amplifies the weakness. A firm's own management information tells it what its existing clients are doing. A well-governed shared ecosystem can begin to show what the wider market is starting to do — but only if the data feeding it is trustworthy first, and the AI sits on top of it second.

A framework I keep coming back to

None of this requires genius. It requires a habit — a short set of questions applied deliberately, before the work begins, that force the frame open before it closes. These are the same five I set out in Part 1 for funding decisions, turned outward to test any proposition that asks a firm to contribute something.

Table 1 — Five questions to interrogate any data or AI proposition before you commit to it

The questions are almost embarrassingly simple. The value is in applying them on purpose, before committing to an answer — and being willing to hand the widened question back to the specialists whose job it is to answer it. Preparation should sharpen the questions you ask. It should not make you certain of the answers before you have earned them.

Prepare hard, then stay genuinely open

There is a failure mode on the other side of all this, and I want to name it honestly. Widening the frame can curdle into arriving with a finished answer and looking for confirmation. That is not thinking; it is rehearsing. The most useful version of preparation produces hypotheses, not conclusions — a better set of questions, held loosely enough to be overturned by the person who knows more than you.

In the meeting that prompted this essay, the reframing I had prepared was not the end of the conversation. It became the starting point for a better one. The people around the table challenged it, extended it and took parts of the thinking further than I could have alone — precisely because I came with a frame rather than a verdict. This is also, in the end, where Jersey's advantage actually lives. The Playbook calls responsible, well-governed adoption “ethical differentiation” — competitive advantage in a jurisdiction built on trust. A shared data or AI utility is one of the clearest tests of that idea: it only works if the firms feeding it believe the exchange is fair, the governance is real, and the intelligence they get back is worth the data they give up.

The next time someone proposes a platform your firm should feed, ask yourself honestly — do you know what intelligence you get back, or are you being asked to answer a question no one has framed properly yet?

Next in this series: why “more information” and “better decisions” are not the same thing — and how to tell which one a dashboard is actually giving you.

Sources: Gartner AI ROI research, code 2026g (80–95% of organisations yet to realise financial AI returns); Jersey AI Playbook (compact-scale advantage, shared infrastructure/centralised KYC utility, “ethical differentiation”); AI Strategy in Jersey FS, Series 1 Part 1 (fund decisions, not tools) and Series 2 Part 4 (data foundations). Illustrative examples are generalised and non-attributable; no proprietary, client-specific, or employer-sensitive information is included.

Professional disclaimer: The views expressed are my own and are shared for general informational purposes. This content is intentionally non-confidential and does not include proprietary, client-specific, or employer-sensitive information. It is not legal, regulatory, investment, or professional advice; readers should seek appropriate advice for their circumstances.

Financial services leader (20+ years) | Lean Six Sigma Green Belt | London School of Economics (LSE) AI Leadership Accelerator

A version of this article was first published on LinkedIn as part of the Sequence body of work. This is the canonical version.

The views expressed are the author's own and shared for general information. This content is intentionally non-confidential and does not include proprietary, client-specific or employer-sensitive information. It is not legal, regulatory, investment or professional advice. © 2026 Paul Soares. Sequence is the trading name of Sequence Advisory Limited (in formation) · hello@sequenceadvisory.com