Guide · AI
How to find AI use cases in your business: a five-step assessment
Most business owners have heard that they should be doing something with AI. The harder question is where to start. According to the US Census Bureau, about one in five US businesses reported using AI in early May 2026. Between December 2025 and May 2026, use grew among firms with 20 or more employees but didn't change significantly among smaller ones. This guide shows how to find AI use cases in your business on your own, in five steps and with a simple scoring table. Some consultants call this an AI opportunity assessment. The goal is the same: find the task where an AI workflow helps first.
What the assessment should give you
A good assessment ends with a decision, not a strategy deck. Three results are enough:
• A list of recurring tasks with a rough estimate of the hours they take each week.
• A score showing which of these tasks suit an AI workflow and which don't.
• A first workflow with a clear goal and a plan to test it with real examples.
Everything else, such as policies, training or a roadmap across departments, builds on that. Teams that start with the roadmap often plan for tasks nobody has looked at closely.
Find AI use cases in five steps
Step 1: List tasks, not ideas
Don't ask "Where could we use AI?" Ask "What do we do over and over, every week?" The first question gets you buzzwords. The second gets you work you can describe. Good places to look are your calendar, your inbox and your task list. Typical answers include replying to inquiries, writing proposals, entering invoices and receipts, writing reports, preparing for and following up on meetings, searching for documents and copying data from one system to another.
Ask the people who do the work. Owners and managers often see only the result, not the twenty minutes of retyping that came before it.
Step 2: Estimate the time roughly
Each task needs two numbers: how often it happens per week and how long it takes on average. Frequency times duration gives you hours per week, for the whole team combined. A rough estimate is enough. You are ranking tasks, not tracking time. If you want more precision, have the team jot it down for one week.
Step 3: Score each task with four questions
Hours alone don't make a good first workflow. Score each task on four more questions, from 1 (low) to 3 (high):
1. How consistent is the task? Does the same kind of input arrive each time, and does the same kind of result come out?
2. How clear are the rules? Could you write down how decisions are made, or does a lot depend on gut feeling?
3. Is the data digital? Does the information arrive by email, PDF or from a system, or on paper and in one person's head?
4. How easily can you fix a mistake? 3 means a mistake is noticed and easy to fix. 1 means it costs money or trust.
Here is what the table can look like. The values are an example, not a measurement:
Replying to website inquiries
Hours per week: 6, Consistent: 3, Clear rules: 3, Digital data: 3, Mistakes easy to fix: 2, Total: 11
Entering packing slips into the inventory system
Hours per week: 4, Consistent: 3, Clear rules: 3, Digital data: 2, Mistakes easy to fix: 2, Total: 10
Writing the weekly project status
Hours per week: 3, Consistent: 2, Clear rules: 2, Digital data: 3, Mistakes easy to fix: 3, Total: 10
Reviewing contracts
Hours per week: 5, Consistent: 1, Clear rules: 1, Digital data: 3, Mistakes easy to fix: 1, Total: 6
Step 4: Pick your first workflow
Choose a task with many hours and a high total. When totals tie, hours decide. In the example above, inquiries come out on top: the most hours and the highest total. Because replies go to customers, this workflow gets a review by a person before anything is sent. Contract review takes a lot of time but would be a poor start: it is different every time, needs a lot of judgment, and a missed clause is expensive.
Don't pick the most impressive task. Pick the one where you will know for sure after four weeks whether it works. Decide right away where a person reviews and approves before anything is sent or posted.
Step 5: Test small, then decide
Before the test, write down what success means, for example "A draft reply is ready, and in most cases it only needs small edits." Then test with 20 to 30 real cases from your day-to-day work, not made-up ones. Track how much time you save and how often a person has to fix the result. Then decide: expand, adjust or drop it. "Drop it" is a useful result too, if you learn it after four weeks instead of a year.
A shortcut for steps 1 and 2
Our Time Sink Check turns frequency and duration for typical tasks into hours per week and shows which task makes a good first workflow. Your inputs are calculated only in your browser and are not stored.
Common mistakes
• Starting with the tool instead of the task. Buying licenses and then looking for uses often ends with licenses nobody opens. More on that in Unused Copilot licenses.
• Starting too big. A workflow across five departments and three systems isn't a test. It's a project.
• Asking about data too late. If the information you need is on paper or in one person's head, solve that first.
• Nobody owns it. A workflow needs a person who looks after it and decides when something changes.
• Overlooking privacy. As soon as customer or employee data is involved, the provider and the contract matter. For a starting point, see Can our team put client data into ChatGPT?.
When outside help pays off
Most teams can handle the first three steps on their own. Outside help pays off mainly when nobody has time for it, when the most promising workflow spans several systems, or when privacy and approvals need to be settled before anything gets built. Our page on AI consulting explains what the work involves. To compare proposals, see How much does AI consulting cost?, and for what to look for in a consultant, see How to choose an AI consultant.
How aenuma helps
If you are looking for a starting point, we review your workflows with you and show which ones are worth automating. The result is a prioritized plan with scope, price and prerequisites. If you already have an idea, we check feasibility, benefit and effort and sharpen it.
Then we build the first workflow in your environment, for example with Microsoft 365, n8n, or Make, and test it with real examples from your day-to-day work. What runs automatically and where a person reviews is agreed in advance. After that, we introduce the workflow to the people who will use it and, if you want, support it afterward.
Frequently asked questions
How long does it take to find AI use cases?
That depends on how many areas you include. For steps 1 to 3 in this guide, plan on one week in which the team takes notes along the way. The test in step 5 runs until you have 20 to 30 real cases behind you.
Do we need an AI strategy first?
Not for the first workflow. A strategy gets better when it builds on experience with real tasks. Only two rules matter from day one: which data may go into which tools, and who approves results.
How is this different from an AI readiness assessment?
A readiness assessment usually asks whether your data, IT and skills are ready for AI. Finding use cases asks where AI pays off. The two are connected. For smaller businesses, the task is the better starting point, because data and IT questions are easier to answer for a concrete case.
Should we do this ourselves or hire someone?
Start yourself if you have the time. The first three steps need knowledge of your business more than knowledge of AI. Bring in help when you have picked a task and need someone to build, test and introduce the workflow.
Sources
Large Firms With at Least 20 Employees Biggest AI Users (US Census Bureau)
By Joshua Schütt, founder of aenuma. Last updated: November 2026.
Where would you start?
Tell us about two or three tasks that take your team time every week. We will tell you which one makes a good first workflow and what it takes.
Book a call