Why Your AI Use-Case Inventory Comes Back Empty

Once the decision to adopt AI is made, the first move is almost always an inventory. Every department gets asked to list the work AI could take over. What comes back is invoice processing, weekly report consolidation, first-line support triage. Named work. Large work. Work with a documented procedure.

Then very little happens, because almost nothing on that list turns out to be worth doing. This is written for people responsible for AI adoption inside an organization. It is also for anyone who has stared at their own job wondering where AI would even fit. The subject is where to look, so it becomes a question about your own week fairly quickly.

That inventory belongs to the previous technology

Look for high-volume work with a fixed procedure. That instruction is not wrong so much as out of date: it is tuned to a different tool. Rule-based automation was genuinely excellent at exactly that. The more identical repetitions there were, the better the return, so listing processes and sorting by volume was the correct method.

Which means the high-volume, highly repeatable work has already had something done to it. What is left over is the work that resisted.

The leftovers are what generative AI is good at. Small in volume, slightly different every time, impossible to reduce to a procedure. Reading something long and working out what it actually asks for. Pulling the useful parts out. Finding the case from two years ago that resembles this one. That work is scattered everywhere and has never been a candidate for automation.

That work has no name

Here is where the inventory breaks. A process register holds work that has a name. "Invoice processing" has a name. "Read the long email from the client, work out what they are actually asking for, and decide who internally should handle it" does not.

With no name, it cannot go on the form. Off the form, it never gets counted. Uncounted, it never becomes a candidate. Meanwhile it happens every day, and the person doing it starts from scratch each time.

This kind of work also varies by person, not just by department. Two people with the same title carry different small judgments. Collect it from the top and that variation is exactly what gets flattened out.

Why "here are some case studies, go find yours" misses

The usual follow-up is a pack of examples from other companies, handed round so everyone can map them onto their own work. The intent is good. But case studies describe named, substantial processes, and translating one of those onto unnamed work of your own takes a step that nobody supplies. That step is left to each reader.

One identical training session for everyone produces the same split. People see the same example, and some carry it back to their desk while others cannot. What separates them is not comprehension. It is the relationship they already have with the tool. That split is where RELACTIS started: a company-wide rollout, one uniform training, and it still did not get used.

What people can spot differs by person

Finding unnamed work is a personal job. Which kinds of it someone notices depends on how they engage with AI in the first place.

RELACTIS is the assessment we built to read that. Twenty-four questions, about four minutes, and you get seven factors (Creation, Crossover, Autonomy, Advocacy, Systemization, Verification, Engagement) plus which of fifteen types you sit closest to. It costs nothing and needs no signup. Take it alone to see your own shape, or run it across a team for the distribution by role.

The differences show up like this.

What often turns out to be short is a role rather than headcount. Three finders, nobody to make anything reusable. With that lineup, good uses keep getting invented and keep stopping with whoever invented them, because there is no seat set up to receive them. Seeing that lets you ask someone to take it, bring someone in, or leave that part narrow for now.

The limits, plainly. What comes out is the shape of the engagement, never the reason underneath it. "Why do I reach for it before anyone tells me to" is a question only that person can answer. Treat the distribution as the opening of that conversation rather than its conclusion.

One thing for tomorrow

Before sending the inventory form round again, think back over your own week and find the small task you rebuilt from scratch even though it barely takes any time. Untangling a long email. Hunting for the previous case. Wording a refusal. Deciding which bucket something belongs in. If it has no name, you are in the right area.

Write one down and you have your search area. Ask five people on your team to do the same and five different things come back. That spread is the material for deciding what happens next.