Cheap at the Top. Expensive at the Bottom.

Monday morning, a data team of three gets an email. There is a new enterprise AI platform, mandatory adoption by Q3, and it replaces the tracker they built two years ago because nothing else answered the question their leadership actually asks them every week. Nobody on that team was in the room when the platform was scoped, and nobody asked what they needed before the requirements were locked. Now they have two jobs instead of one: keep the mission running, and figure out how to make the mandate work, or quietly build around it again.

That scene is not hypothetical. It is the default outcome in many large bureaucratic organizations right now, and it shows up as a resourcing pattern as much as a design failure. The teams chartered to build the enterprise platform, the model, or the dashboard sit high enough in the org chart to be large and well-funded. The teams who actually have to run the thing sit low enough to be thin, already maxed out on their existing workload, and last on the list of people consulted before the solution ships.

Leadership and the customer keep getting treated as the same person, but they rarely are. We previously described what happens when that false equivalence goes unchecked: a solution calibrated to the questions leadership is asking, built on data that nobody at the operational level trusts, with an open invitation to walk the design team through the reality on the ground that is never taken up.

Give that misalignment enough time and the workaround always shows up. Why Subordinate Organizations Keep Building Excel Trackers: named the reason:

Enterprise systems optimize for scale and auditability; operators need speed and fit for the task in front of them; and when the official system doesn’t close that gap, someone closes it themselves.

It used to be a spreadsheet on a shared drive. Now it’s an unsanctioned agent, a personal prompt library, a script somebody wired up to a model API over a weekend. Same instinct, sharper edges: a rogue spreadsheet duplicated a number, an unapproved AI tool can leak data, hallucinate a decision input, or quietly become the thing an entire team actually relies on, unsupported and ungoverned, because nobody with the authority to fix the real system ever asked what was missing.

None of this is a discipline problem. The tracker isn’t the rebellion; it’s the diagnosis. The workaround is the rational move of someone still trying to get the job done inside a system that wasn’t built for them. What we haven’t addressed is:

For a well-staffed team, building a bridge solution is an inconvenience. For a three-person shop already at full stretch, it’s a second job layered on top of the first, done on nights and weekends, with none of the governance or support the official system at least pretends to offer.

Two failures have carried this series so far: a discovery gap and a trust gap. There is a third one hiding underneath both, and it rarely makes it into the requirements document. Let’s call it the capacity gap.

The level that skips discovery does not pay for skipping it. It has the staff to absorb a design miss, run another governance review, tolerate a slipped rollout. The level below it has none of that room, and the cost of the mistake doesn’t disappear. It rolls downhill and lands on exactly the people least equipped to catch it, which is how a miss that registers as a rounding error at the top can eat a small team’s entire month.

The fix starts with a question discovery interviews rarely ask: not just what does the user need, but what will they have to stop doing to make room for what you’re about to give them, and did anyone budget for that? Skip it, and all three gaps compound at once: discovery, trust, and capacity, each one absorbed furthest from the level that made the decision.

If your organization is watching under-resourced teams get pulled off their real work to implement or work around a top-down data or AI initiative, Storm King Analytics is glad to help you trace it back to which of these gaps is driving it. Reach out at info@stormkinganalytics.com.