You keep hearing that AI could help your business, and you believe it. The part that stalls people is smaller and more practical: which task do you hand it first? Sitting down to a blank chat box with no plan is where the overwhelm starts. This walkthrough gives you a fifteen-minute task audit that turns "AI, somewhere, somehow" into three specific things to try this week. Spotting where AI actually fits your work is one of the early skills in The 7 Levels of AI Proficiency, and you can practice it today with a pen and a timer.
What a use case actually is
A use case is one repeated task you already do, matched to a result you can describe. It is more specific than a general "use AI for marketing." Think about drafting a first-pass reply to a customer question that you would actually send. The tasks that fit best early on are predictable and repeated, with structured inputs and outputs: numbers, dates, names, statuses, standard replies. Predictability is what makes a task a good candidate, because you can tell quickly whether the result is right.
There is data behind starting task-first. In a May 2025 survey of 947 small businesses with annual revenue between $25,000 and $5,000,000, 34% of owners said they did not see a clear use case or return, and 37% said they lacked the time or resources to explore tools. Those two are the same wall: no obvious place to point it.
of small business owners said they did not see a clear use case or return from AI. The other common wall, at 37%, was lacking the time or resources to explore tools.
Source: PayPal / Reimagine Main Street Small Business AI Survey, 2025When it is worth doing
You do not need every task to be a fit. You need one. A task earns a spot on your short list when it checks most of these:
- You repeat it every week rather than once a quarter.
- The inputs are structured (a form, an email, a spreadsheet row, a standard question).
- A wrong answer is low-stakes and easy to check before it goes anywhere.
- It eats real time, so a faster draft is worth something.
Marketing and customer engagement is a common starting point: in that survey, 77% of active AI users named it the area where new AI solutions would have the greatest impact. That does not mean it has to be yours. The point is to choose from your own repeated work rather than a tool's feature list.
You do not need every task to be a fit. You need one.
The task audit, step by step
The whole thing runs on a timer and a list. Work through these six steps in order, and finish with three tasks to try this week.
- List the work you repeat first, tools later. Set a timer for fifteen minutes. Write down every task you repeat in a normal week. Skip the AI question for now. Next to each one, estimate the minutes per week it takes you.
- Circle the most repetitive, time-consuming, or annoying ones. The strongest first candidates are the tasks you dread because they come back constantly and follow a pattern. High repetition plus structured inputs is the signal.
- Define what a good result looks like before you try anything. For your top pick, write one sentence describing a result you would actually use. Anthropic's own guidance says a good criterion is Specific, Measurable, Achievable, and Relevant. Aiming to draft a reply a customer would actually send beats aiming to help with email.
- Rate each task's AI potential. Score how repetitive it is, how structured the inputs are, whether a wrong answer is low-stakes and easy to check, and how much time it eats. High repetition, structured, low-stakes, high-minutes goes to the top of your list.
- Keep yourself in the loop. Treat the first week as a test, not a handoff. Read the first outputs against the sentence you wrote in step three, and approve each one before it goes out.
- Pick three to try this week. Rank your list and take the top three. Three small tests beat one big automation project that stalls before it begins.
Do it now
Once you have your list, the audit itself takes a few minutes with any assistant. Paste your tasks into the prompt below and let it score them with you. It will not know your business better than you do, so read its ranking as a starting draft rather than a verdict.
I want to find my first AI use cases. I am going to paste a list of tasks I repeat every week, with a rough estimate of minutes per week for each.
For each task, rate it 1 to 5 on four things:
1. How repetitive and predictable it is
2. How structured the inputs are (forms, emails, numbers, standard questions)
3. How low-stakes a wrong answer is, and how easy it is to check
4. How much time it costs me each week
Then give me a ranked short list of the top three tasks to try first. For each of the three, write one Specific, Measurable, Achievable, Relevant sentence describing what a good result would look like. Ask me one clarifying question before you rank if anything is unclear.
Here is my list:
[paste your tasks and weekly minutes here]After it ranks your tasks, run this shorter prompt on the number-one pick to pressure-test the result before you rely on it.
Here is the task I want to start with: [paste the task].
Here is my definition of a good result: [paste your sentence].
Draft the result once. Then critique your own draft against my definition, name anything that would stop me from using it as-is, and produce a revised version. Run it as closely to my definition as possible.A starter file to download
To make the fifteen minutes easier, there is a starter worksheet you can fill in. It has a column for each task, a column for weekly minutes, the four rating columns from step four, and a spot for your one-sentence definition of a good result. A worked example is filled in at the top so you can see the shape before you start. Rename it, clear the example, and put your own tasks in.
Download: First AI Use Cases Starter Worksheet (how-to-find-your-first-ai-use-cases-starter.md). A real template you fill in, no signup.
Common mistakes to skip
A few patterns slow people down, and you can step around all three.
Starting from the tool instead of the task. Buying a subscription first leaves you staring at a blank box wondering what to type. The list comes before the login.
Skipping the definition of a good result. This is the step that turns an assistant into a useful teammate. Without a target sentence, you cannot tell whether the output is right, so every result feels vaguely off. Write the sentence first.
Trying to automate the whole thing on day one. Keep yourself in the loop early. Read the drafts, approve them, and learn where the assistant is strong and where it needs a tighter instruction. As Anthropic describes the model for its small-business tools, "You approve the plan first or, when you're ready, let it run end-to-end." Ready comes after a few good weeks, not on the first afternoon.
Where this fits in the bigger picture
Finding one task and handing it to an assistant looks small. It is the on-ramp to a durable skill. Once you can spot the repeated, structured work in front of you, describe a good result, and check the output, you can do it for the next task, and the one after that. That is the difference between using AI as a novelty and building a system that runs a real part of your week.
The functions where this compounds are the same six that small businesses target most: finance, operations, sales, marketing, HR, and customer service. You do not have to touch all of them. Pick the one task that annoys you most this week and start there. The survey backs the reason to begin: 74% of owners said they would adopt AI with clearer evidence of what it returns, and a completed task on your own desk is the clearest evidence you can get.
If you want to see where your own skills sit today, The 7 Levels of AI Proficiency assessment takes about ten minutes and gives you a read plus a next step.
Related reading: Level 5: The Captain (Design Thinker).
Sources
- Claude Platform Docs: Defining Success Criteria for LLM Applications
- PayPal / Reimagine Main Street Small Business AI Survey (fielded May 2025, n=947; released June 10, 2025)
- Anthropic: Introducing Claude for Small Business (May 13, 2026)
- The 7 Levels of AI Proficiency assessment
Frequently Asked Questions
What is an AI use case?
A use case is one task you repeat, paired with a result you can describe. Instead of a general "use AI for marketing," it is drafting a first-pass reply you would actually send. The clearer the task and the target result, the easier it is to tell whether the AI helped.
Do I need to know how to code to start?
No. The task audit uses a timer and a list. The prompts are copy-paste, and you read the results in plain language and approve them yourself.
How do I pick my first task?
Choose the one you repeat most often that has structured inputs and a low-stakes, easy-to-check output. Repetitive, predictable work is the strongest starting point.
What is the difference between a task and a full automation?
A task is one repeated job you can hand over and check each time. A full automation runs end to end without you watching each result. Start with the task, keep yourself in the loop, and let it run on its own only once it earns your trust.
Does this only work with one AI tool?
No. The audit is about your work, not a product. Any capable assistant can run the scoring and drafting prompts. The six functions where this most often starts, finance, operations, sales, marketing, HR, and customer service, are common across tools.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.