AI readiness assessment for small businesses: a 10-minute checklist

What a small-business readiness check should measure

Enterprise readiness frameworks score data warehouses, AI policies, model governance, and hiring plans. Those questions matter for a company with hundreds of employees. For a business with two to fifty people, they mostly produce a long report and no decision.

The useful question is narrower: do you have one repeated piece of work, with information you can reach, a person who will own the result, and enough tolerance for a supervised test? If the answer is yes, you are ready to try something small. If not, the checklist below shows exactly what to fix first.

• The work: is there a repeated task with a recognizable pattern?
• The information: can the inputs be reached without heroic effort?
• The people: will someone own the test and review the output?
• The risk: what happens when the result is wrong?
• The baseline: can you measure today’s version of the task?
• The appetite: will the team use a new process if it works?

Score the work: repetition and clarity

Give one point for each statement that is true for the task you have in mind. Be specific—“customer emails” is too broad; “answering quote requests that arrive through the website form” is specific enough to score.

Illustrative example: a four-person property management office answers the same handful of tenant questions dozens of times a week. Those questions repeat, follow known rules, and already have approved answers in a policy binder. That task would score well here, even though the office has never used AI.

• The task happens at least weekly
• Two people would describe the steps the same way
• You can show a recent example of a good result
• Unusual cases are recognizable when they appear

Do this: If this section scores below 2, pick a different task. A vague or rare task is the most common reason a first AI project stalls.

Score the information and the people

AI and automation can only use what they can reach. If the answer to a customer question lives in one employee’s memory, the first step is writing it down, not buying software. See what belongs in a business knowledge base for a practical way to do that.

People matter just as much. A small pilot needs one named owner who reviews output, notices problems, and decides whether the test continues. Without that person, problems go unnoticed until a customer finds them.

• The inputs live in a system you can export or connect (email, forms, a spreadsheet, a CRM)
• Approved answers, prices, or rules are written down somewhere
• One person can spend an hour or two a week reviewing results during a test
• The people who do the task today are willing to try a new way

Score the risk and the baseline

Readiness is not only about capability. A task where a wrong answer costs a refund, a safety problem, or a legal issue needs much more care than one where a wrong draft is simply corrected before sending.

You also need a “before” number. Without it, you cannot tell whether the change helped or just felt busy.

• A mistake can be caught by a person before it reaches a customer
• The task does not involve sensitive health, legal, or financial decisions
• You can count how often the task happens and roughly how long it takes
• You could keep the old process running as a fallback during a test

Read your score and choose a next step

Add up your points across the four scoring sections (sixteen possible).

• 12–16: Ready for a small pilot. Pick one tool or automation, define a stop rule, and test for two to four weeks. The small AI pilot guide walks through it.
• 7–11: Close. Fix the lowest-scoring section first—usually missing written answers or no clear owner—then rescore.
• 0–6: Not yet, and that is useful to know. Choose a more repeatable task or start by documenting the process. The guide to finding the first AI opportunity can help you choose.

Do this: Write your score, the weakest factor, and the one fix you will make this month at the top of the scorecard below.