vibencode
Use-case strategy
Which problem should we point AI at first?
“The board wants an AI roadmap.”
Every candidate use case assessed for value, feasibility and risk, ranked against each other, with one pilot brief specific enough to start on Monday.
The list is always longer than anyone expects, and it is never the list that was in the deck. Four of the entries turn out to be one idea described by four departments who have not spoken to each other; a third of them die the moment somebody looks at whether the data exists. That is the work: not ranking a shortlist somebody already wrote, but finding out what is genuinely on the list, then scoring every candidate the same way so a disagreement has something specific to argue with. Most of the list below is that.
- 01Work the candidates through with the people who own the processes, not only the people who own the budget.
- 02Assess each for value, feasibility and risk, including the ones we will recommend against.
- 03Rank them, and say plainly why the first beat the second.
- 04Write one pilot brief specific enough to start on Monday.
- 05Leave the discarded options legible, so the decision survives a change of sponsor.
- Opportunity assessmentEvery candidate we found, including the ones nobody proposed and the ones we recommend against.
- Feasibility reviewWhat the data can actually support, established by looking at it rather than by asking.
- Prioritised pilot briefOne candidate, scoped tightly enough that a team could start on Monday without us.
Four of the nineteen were one idea, described by four departments who had not spoken.
Gather
Interviews across functions. The list is always longer than anyone expected, and that is the point.
Assess
Value, feasibility and risk, worked through with the people who own the process.
Rank
Scored the same way, reasoning attached, disagreements surfaced rather than smoothed over.
Brief
One pilot, specific enough that a team could start on Monday without us.
That is the shape of a full engagement. If it is more than you want to commit to yet, the same people will spend two days on one piece of it first, and you are free to stop there. Start a 48-hour sprint
- Process owners, not only sponsorsThe people who do the work know which step is painful. Budget holders know which step is visible. Both are useful; only one is enough.
- An honest view of the dataIncluding what is not collected. Half the candidates die here, and it is far better they die in week two than in month seven.
- A real constraint on time and money“Unlimited” produces a roadmap nobody funds. Tell us the real envelope and the ranking becomes a decision rather than a wish list.
- Agreement on who decidesA ranking with no owner becomes a document. We need the person who will stand behind the choice in the room while it is made.
What actually is use-case strategy?
Finding every place AI could plausibly help in your organisation, assessing each one for value, feasibility and risk on the same scale, and ranking them with the reasoning attached. What you get is a decision somebody can defend, including the candidates we recommend against and why, plus one pilot written tightly enough that a team could start on it without us.
How long does this take?
Typically four weeks for a mid-sized organisation. Faster is possible and usually means fewer interviews, which is precisely where the value of the exercise sits.
What if the answer is “none of these”?
Then we say so, and four weeks has saved you a year. It has happened, and from where we sit it is a good outcome rather than an embarrassing one.
We already have a shortlist. Can you just validate it?
Yes, although we will ask what is not on it. The most common finding is that the strongest candidate was never proposed, because it sits between two departments and belongs to neither.
Do you then build the pilot?
Often, but not necessarily, and the brief is deliberately written so that anyone could. If we were the only firm able to execute it, it would be a bad brief.
Is this a workshop?
There are sessions in it, but the deliverable is a ranked assessment and a pilot brief, not a set of photographed sticky notes and a summary email.
If you already know the problem and the open question is whether a given approach will hold up, that is a technical investigation rather than a prioritisation exercise. That is applied research.
Which problem should we point AI at first?
Two days on the list you already have. Written up at the end.