“Can you pull the new FTS numbers?” The message looks like a thirty-second favour. The first part of this pair followed it downstream, into a wrong number baked into an expensive decision, and landed on a single observation: the recurring failure is a question that was never framed properly before the work began.
The fix is smaller than it sounds. No reorganisation required. Two mirror habits, one on each side of the table, and a structure leadership builds around them.
State the decision
“Pull the FTS numbers” is a request for output. “I am deciding whether to renew FTS for another season, and I think completion rate on the new season is how we will judge it” is a request for help with a decision. The second version quietly answers most of the hidden questions on its own. Metric, period, comparison, and purpose all fall out of the decision once it is named.
So the business side of the bargain is a one-sentence habit: name the decision before asking for the number.
The same rule holds the moment a chat window opens with a language model. A vague prompt is the same mistake as a vague request, with one difference. The model will not push back.
Ask one question first
The data side commits to a habit too. When the next vague ask arrives, do not pull a number. Ask one question first: “What decision is this feeding?”
That single question heads off a surprising amount of rework, and a surprising amount of false confidence further down the line. It is also how a data team earns back the trust the bad numbers quietly spent.
And when a metric has no shared definition, do not pick one in private. Write the definition down, get it agreed, and put it where the next person will find it. One blessed definition per metric is what stops three teams computing “views” three different ways.
What leadership owes
Goodwill doesn’t scale. Two colleagues can settle the FTS question over coffee; a large organisation needs structure. Building that structure is leadership’s job, and the part most often skipped.
A named bridging role, for one. Someone who turns a business problem into a solvable data problem, and the answer back into a decision. It can be filled from either side: an analyst with business sense, or a business person literate enough to frame the question and read the answer.
Clear ownership of metrics, for another. One team accountable for a given metric’s definition and quality, so that when “views” is wrong, a name is attached to fixing it. What fails is the arrangement where nobody owns the metric on either side.
And hiring for data literacy beyond the data team: marketing, product, operations, finance, HR. Hiring criteria signal what an organisation values. If the criteria for those roles never mention data, leadership has decided literacy is somebody else’s problem.
Democratisation, two ways
“Data democratisation” gets sold as an unalloyed good. It comes in two versions that share a word and nothing else.
Done badly, it hands people raw access and calls it empowerment. Done well, it pairs access with the things that make access safe: shared definitions, a small set of trusted standard reports, and the literacy to read them. The difference between the two is governance, not ambition.
AI amplifies the foundation
AI changes none of this. It amplifies whatever the foundation already supports. Give it shared definitions and a literate workforce and it helps. Give it the mess, and it returns the mess with more polish and more apparent authority.
The foundation is the part an organisation actually controls. The model is not.
Where to start
Pick one thing for Monday.
Making requests? Stop sending “pull the FTS numbers.” Send the decision instead.
Fielding them? Ask “what decision is this feeding?” before pulling anything, and write down the next metric definition rather than guessing at it.
Leading the people doing both? Fund a named bridging role rather than hoping someone volunteers, and decide what the organisation is shipping, governed self-serve or raw access, because AI will scale up whichever one is chosen.
None of it costs a new platform. What it buys back is trust in the numbers, which is slow to rebuild and worth more than any dashboard. The next FTS request arrives tomorrow. Whether it stays a thirty-second favour or opens a real conversation depends on which habits got built first.
Part one lays out the problem; the full argument, with sources, lives in the long-form article.