The builder
This one is written for the leader, not the builder, because the builder did the thing leaders say they want. Someone in sales operations at a shipping company, not an engineer, with a real problem the official systems did not solve: the team could not see at a glance which vessels were expected at which ports. So they described the app to a platform that builds apps from descriptions, and by the end of the day the team had it.
That is internal talent worth developing, seen from the inside. The initiative, the tools, and the judgement to solve a real problem were already on the payroll. What was not on the payroll was a decision about what any of it was allowed to touch.
What the AI made easy
The app. An afternoon, no ticket, no budget line, and a working tool the team actually used. That is the new fact of the market, and it is a good one.
What the platform made equally easy was the rest. Its default for a new app was public. The database key sat in the front end, as it does on that platform, safe only if the access rules are set, and nobody set them. The vessel schedule, which is operationally sensitive in an industry where knowing which ship is where and when is worth money to competitors and worse to others, was readable by anyone who found the address. In May 2026 a security firm looking for unauthorized AI use found roughly five thousand employee-built apps like it exposing corporate data, built by product managers, analysts, and other people outside engineering, on platforms that default to public: shipping schedules, a health company's clinical trials, a bank's records, a care facility's patient conversations. Reporters verified several, this one among them. The company learned it was exposed from the press.
The analyst had done the right thing badly. The leader who answered for it had never been asked to decide anything.
The check
Every resource runs the same three rows. Founder judgement decides the first. Engineering judgement decides the third. Observation connects them.
| Row | Verdict | Evidence |
|---|---|---|
| Originated itShould this exist, in this form? | surrendered | Should this exist, on this data, in this place? Nobody in the company had decided. The analyst decided alone, because there was no one to ask and no rule to read. That is a founder-judgement gap, and the founder here is the leader. |
| Observed itWas anyone reading the work while the AI produced it? | surrendered | Nobody read it, because nobody knew it existed. The platform's defaults were the only review the app ever got. |
| Answered for itWill it hold, and who stands behind it when it ships? | surrendered | No name was on it until the reporter called. Then the name was the company's. |
Four flaws, every one of them the leader's to fix, not the analyst's:
- Talent without a gate. The initiative is the asset. With no permission structure around it, the talent works in the dark, and the dark is where the defaults win.
- The platform's default as company policy. When nobody has written down what may be public, the vendor's checkbox has decided it for you.
- Data leaving the building unnoticed. Operational records went to a third-party platform and out to the internet in one afternoon, and the company's security process never saw a request, because there was nothing to request.
- Blaming the builder. The reflex after a story like this is to discipline the one person who built something. That punishes the exact behavior the company needs and teaches everyone else to hide theirs.
What judgement would have decided
What an experienced leader would have had in place before this analyst ever opened the platform:
- FounderDecide, in writing, what anyone may build and on what data. Three tiers are enough: data that may be public, data that stays inside, data that is regulated and never leaves a named system. One page, and a conversation with every person who has ever built a spreadsheet macro.Costs: an afternoon, and a few uncomfortable conversations about what is already out there.
- FounderDevelop the builder instead of disciplining them. This is the person you have been saying you want. Give them the standing to build, a mentor, and the gate to build inside. The alternative is losing them to a company that will.Costs: a mentor's time and, eventually, a title.
- EngineeringA default for every internal build. Private by default, credentials in the company's name, and a short checklist that runs before real data enters: who can see this, what happens if the address leaks, who is on it. The platform will not do this for you.Costs: a slower first afternoon, every time.
- EngineeringInventory what already exists. Scan for what employees have already built and published. The count will be higher than anyone expects, and some of it will be worth keeping.Costs: a day, and the findings.
- EngineeringA named answerer for each internal tool. The builder, with a second name beside it. The name is what turns a shadow app into a company asset, and it costs nothing but honesty.Costs: nothing, except the honesty.
Where the fast path was right
The analyst building it. That afternoon is the whole point of the internal-talent situation: a real problem solved by a person who already understood it, without a ticket and without a vendor. Nothing above takes that away. Private by default and a name on the tool would have cost the analyst an hour and changed nothing about the app.
What changes is where the risk sits. It moves from the open internet to the leader's calendar, as a policy to write, a person to develop, and an inventory to face. That is the trade: an afternoon of governance and a mentor's attention, in exchange for never learning from a reporter what your own people built.
On the public record, May 2026, names removed. Written with AI assistance; the judgement is mine. — Clay