How to Identify AI Use Cases in Your Business
To identify AI use cases in your business, stop asking people where AI could help and map how one important workflow actually runs today. Walk it stage by stage with the people who do the work, mark each stage as judgment, coordination or waiting, and look for the coordination stages that repeat dozens or hundreds of times a week. Those are your best candidates.
Most guides on this topic hand you a brainstorm and a two-by-two matrix. The matrix is fine. The brainstorm is the problem. This post gives you a method you can run in a week with one workflow, a whiteboard and two or three of your people.
Why does "where can we use AI?" produce weak answers?
Because people answer with tools, not work. Ask a leadership team where AI fits and you get the demo someone saw last week, a chatbot for the website, or "something with our data". None of these starts from a problem anyone has measured.
Who you ask matters too. Managers describe the process as it was designed. The people doing it know where it really breaks: the spreadsheet nobody else has seen, the email that gets retyped into another system, the follow-up call that happens every Tuesday. RAND's study of failed AI projects lists five root causes, and the first is that stakeholders misunderstand or miscommunicate what problem AI is supposed to solve. A brainstorm in the boardroom is a fast way to make that mistake.
Where do the best AI use cases usually hide?
In the work about the work. Two terms help here. Judgment work is what customers pay you for: pricing a risk, diagnosing a problem, deciding whether to take on a client. Coordination work is everything that moves judgment work along: keying data from emails, chasing documents, formatting reports, routing requests, checking one system against another.
Coordination takes a bigger share of the day than most CEOs assume. Asana's Anatomy of Work Index, a survey of 10,624 knowledge workers, found people spend 58% of their workday on work about work and less than 10% on strategic work. When you draw a workflow out, the expensive experts often sit in a handful of judgment stages, while most of the stages around them are coordination.
That is also where the returns tend to be. MIT's NANDA research found that more than half of generative AI budgets go to sales and marketing tools, while the biggest ROI showed up in back-office automation. Budgets follow what is visible. Value sits in what is tedious.
How do you find AI use cases? A five-step mapping method
Budget about a week for your first workflow: one session to map it, a few days to collect numbers, and one session to review what you found.
- Pick one workflow that moves a number. Choose something with a clear start and end that you already measure: inbound request to sent quote, referral to first appointment, order to invoice, month-end close. If you cannot name the metric, pick a different workflow.
- Walk it with the people who do it. Book 60 to 90 minutes with two or three people who touch the work daily, plus their manager. Start at the trigger and list every stage until the work is done. For each stage, write who does it, which system they use, what comes in and what goes out. Keep asking "then what happens?" until the answer is "it's finished".
- Label every stage J, C or W. J is judgment: a decision that needs expertise. C is coordination: moving, copying, checking, formatting or chasing. W is waiting: the work sits until someone or something responds. Be strict. A step called "review" is often coordination with a nicer name.
- Put a number on each C and W stage. How many times a week does it happen, and how many minutes does it take? Pull counts from your systems where you can, and ask the team for honest estimates where you can't. Then multiply. A 10-minute stage that runs 300 times a week is 50 hours of work, more than one full-time person. For waits, note how many days they add for the customer.
- Write each candidate as a What. One sentence with the job, the team and the number that will show it worked. "Cut the time from referral received to referral entered from two days to two hours" is a candidate. "AI for intake" is not.
A single workflow will usually surface several candidates. Map two or three workflows and you have a better list than any brainstorm, with a number already attached to every idea.
What patterns should you look for on the map?
Once the stages are on the wall, four patterns are worth hunting for. Each is coordination work that current AI tools handle well, with a person checking the output.
- Email as the intake system. Requests arrive as emails, PDFs or web forms, and someone retypes them into your CRM or job system. AI can read the message, extract the fields and fill the record for a person to confirm.
- The same step done several times. A document is prepared, split, checked and checked again by different people in different teams, often in the same software. The map makes the repetition visible. Sometimes the fix is one owner, not AI.
- Chasing as a job. Follow-ups for missing documents, signatures or approvals run on a schedule. AI can track what is outstanding, draft the reminder and flag the cases that need a human call.
- Rebuilding from scratch. Every proposal, report or customer setup starts from a blank page or a messy old file. AI can assemble a first draft from your templates and past work, so the expert edits instead of writes.
Watch for one more: long documents that people read only to pull out a few facts. Turning a 40-page packet into a one-page summary with page references is often a quick win, because the reader can check every line against the source.
What should stay with your people?
The J stages. AI should prepare judgment, not replace it. Give the expert a summary, the relevant history and a flag on anything unusual, then let them decide. In regulated work, a human decision is often required anyway.
The evidence says this kind of help pays. In a study of 5,179 customer support agents, Stanford and MIT researchers found an AI assistant raised issues resolved per hour by 14% on average, and by 34% for novice and lower-skilled workers. The assistant suggested, and the agent decided. That is the pattern to copy for judgment stages: faster and more consistent, still owned by a person.
How does mapping compare with brainstorming use cases?
- Starting point: a brainstorm starts from what AI can do. A map starts from how the work runs today.
- Who talks: a brainstorm hears from leaders. A map hears from the people who do the work.
- What you get: a brainstorm gives you ideas. A map gives you candidates with a frequency, a time cost and a metric.
- Typical first project: a brainstorm picks the most impressive demo. A map picks the most repeated coordination stage.
- Side benefit: a map often shows fixes that need no AI at all, like removing a duplicate step or giving a stage one owner.
That last point matters more than it sounds. Put AI on a messy process and you get the same mess, only faster.
What happens after you have the list?
You rank it. Score each candidate on impact, effort and readiness, then pick no more than three. We cover the scoring step by step in How Should a CEO Prioritize AI Projects?
The map keeps paying off after that. At Journey we use a simple order for any AI idea: What, How, Build. The What is the job and the number, which step five gives you. The How is the future workflow: which stages AI takes on and where a person reviews. Your map is the first draft of the How. Only then do you choose whether to configure, buy or build, which we explain in Build or Buy AI? How a CEO Should Decide.
This order lines up with what separates companies that get results. In McKinsey's 2026 State of AI survey of 1,719 respondents, nearly three-quarters of AI high performers had fundamentally redesigned workflows because of AI, compared with about a quarter of other respondents. You cannot redesign a workflow you have never drawn.
What should you do this month?
- Name one workflow that touches a number you report every month.
- Book 90 minutes with the people who run it. Map every stage and label it J, C or W.
- Count frequency and time for the coordination stages, and write the top three as a What.
- Share the map with your leadership team before anyone mentions a tool.
If you want a second pair of eyes on where AI fits in your company, Journey offers a free CEO AI audit. Either way, start with the map. Your best use cases are probably already on it, sitting between the decisions.
Frequently Asked Questions
Map one workflow that affects a number you already track. Walk it with the people who do the work, list every stage, and label each one as judgment, coordination or waiting. Count how often each coordination stage happens and how long it takes. The stages that repeat most and cost the most time are your strongest AI use cases.
Good first use cases are frequent coordination tasks with a person checking the output: turning emailed requests into system records, summarizing long documents, drafting proposals or reports from templates, and tracking follow-ups for missing documents or approvals. They tend to use tools you already have, show results within weeks and build trust before larger projects.
Involve the people who do the work every day, plus their manager. Frontline staff know where the process really breaks, such as retyped emails, private spreadsheets and repeated checks. Leaders should join to confirm which metric matters and to own the result, but a list built only by executives tends to start from tools rather than real problems.
Plan on about a week per workflow: one 60 to 90 minute session to map the stages, a few days to collect counts and time estimates, and one session to review the candidates. Mapping two or three workflows over a month is usually enough for a solid list you can rank and act on.
Judgment work is the expert decision your customers pay for, such as pricing a risk or diagnosing a problem. Coordination work moves that decision along: entering data, chasing documents, formatting, routing and checking. AI usually pays off fastest on coordination work, while for judgment work it should prepare information so a person can decide faster.
