How to Run an AI Use Case Discovery Across Your Company

Most companies pick their first AI projects in a meeting. The people who feel the pain every day are rarely in it. Here is a practical way to find the AI use cases worth building, and to rank them before you spend anything.

What is an AI use case discovery?

An AI use case discovery is a structured effort to find where AI would remove real work in your company: the repeated, manual, error-prone tasks that people put up with. The output is not a list of ideas. It is a ranked list of specific, buildable tools, each with an estimate of demand, time saved and cost.

Why the usual approaches fall short

• Leadership brainstorms reflect what leadership sees, which is rarely the daily friction on the ground.
• Company-wide surveys produce vague answers such as "better search" or "less admin", and low response rates.
• Workshops favour the loudest voices and cover a handful of teams at best.
• Vendor pitches start from the product, not from your problems.

A five-step process

1. Define the scope. Decide which teams and processes are in. Starting with the whole company is possible, but tell people why you are asking and what will happen with the answers.
2. Interview, don't survey. Ask each person about a recent week: what took the most time, what they repeated, what they dreaded. Follow up on specifics ("how often?", "who else does this?"). Specifics are what make an idea buildable.
3. Capture ideas in one format. For every idea record the task, the team, how often it happens, how long it takes, and what a good result looks like. One format makes ideas comparable.
4. Cluster duplicates. Twenty people describing the same pain in different words is one use case with strong demand, not twenty ideas. Merging them is where most of the signal comes from.
5. Rank by demand, value and cost. Score each cluster on how many people want it, the hours it would save, and a rough build cost. Fund from the top of the list.

What a good use case looks like

"Use AI for reporting" is a theme. A use case is specific: "build the weekly executive deck from source data and flag numbers that look wrong before it goes out." It names a task, an owner, an input and a success condition. If you can't say who would use it on Monday, it is not ready to build.

Common mistakes

• Treating the most requested idea as the most valuable. Demand and value are different columns.
• Ignoring teams that don't speak up. Quiet teams often have the clearest repetitive work.
• Skipping feedback. People who see their idea on the roadmap adopt the tool. People who never hear back stop contributing.
• Going straight to a model or vendor before the problem is defined.

Doing it at scale

The method above works with ten people. With a thousand, the interviews and the clustering become the bottleneck. This is the problem Kogi is built for: an AI agent that holds the one-on-one conversations inside Slack or Teams, asks the follow-up questions, and groups the results into a ranked, costed roadmap.

Frequently asked questions

What is an AI use case discovery?

A structured effort to find where AI would remove real work in a company. The output is a ranked list of specific, buildable tools, each with an estimate of demand, time saved and cost, not just a list of ideas.

How long does an AI use case discovery take?

With a handful of teams, a few weeks of interviews and clustering. Done by hand it scales poorly, because interviewing and grouping answers become the bottleneck as headcount grows. An AI interviewer such as Kogi can run all the conversations in parallel.

Why interview employees instead of sending a survey?

Surveys produce vague answers such as better search or less admin. Interviews let you follow up on how often a task happens, who else does it and what a good result looks like, and that detail is what makes an idea buildable.

How do you rank AI use cases?

Cluster duplicate requests first, then score each cluster on how many people want it, the hours it would save and a rough build cost. Demand and value are separate: the most requested idea is not always the most valuable.

What is the difference between a theme and a use case?

A theme is broad, such as use AI for reporting. A use case names a task, an owner, an input and a success condition, such as building the weekly executive deck from source data and flagging numbers that look wrong.