- Asked whose objective improves if this works, the honest answer is a team, or nobody.
- Nobody has written down the result it has to reach, or a date for deciding.
- You could not say what the task it replaces costs you today.
“Leaders must own everything in their world. There is no one else to blame.”Jocko Willink and Leif Babin, Extreme Ownership, St. Martin's Press
What this phase buys you
Progress becomes measurable. Someone is accountable for the outcome, and there is a number describing how things stand today.
The owner, today's number, the decision that changes when this works, and the date you decide. Four lines, one page, signed by the person named on it. If it needs a second page, the phase is not done, it has been widened.
Not this. Not an org chart and not a RACI. Both describe who is involved. Neither names who is answerable, and a RACI in particular spreads accountability across enough boxes that none of them carries it. Stewardship, in the DAMA sense of a named owner per data source, comes in Find it and Agree it.
This phase is about the business outcome, not the data.
One named person's objective improves if this works, and you can state today's cost or cycle time as a figure rather than an impression.
Every later phase produces work nobody is answerable for, and no result can be accepted because there is nothing to compare it against.
What you actually do
Four moves. None of them needs a platform, and none of them is glamorous. That is the point: this is the work that gets skipped, which is why so many companies are stuck a phase earlier than they think.
Name one person, not a committee
Find whoever already carries the outcome this touches and make the result part of their objectives. Not a steering group. One person who can also say stop.
Measure two weeks of the existing process
Before changing anything, record what the current way costs. Hours, cycle time, error rate, rework. Two weeks of unglamorous numbers.
Write the decision that would change
Name the specific decision or action that happens differently once this works. If no decision changes, there is nothing to build.
Set a go or no go date
Pick a date and a threshold in advance. A pilot with no decision date does not end, it just gets quieter.
The usual mistake
Buying a tool and hoping ownership emerges from enthusiasm. It does not. Enthusiasm decays on a predictable curve and leaves the licence behind. Willink and Babin put the same point about standards more usefully: it is not what you preach, it is what you tolerate. A company that says the result matters and tolerates nobody being answerable for it has already told you which one it means.
- No decision owner. Nobody's objective moves when the AI works. The output exists and no one is accountable for acting on it, so it is admired in a meeting and changes nothing.
- No baseline to improve on. The current process was never measured, so better cannot be demonstrated. Any result can be argued with, which means no result is ever accepted.
- The tool arrived before the problem. A platform or licence was bought and a use case is being sought for it. The question being answered is what can this do, rather than what is expensive here.