Is your SME ready for AI? Assess people, processes and data

A small business does not need a dedicated AI department to start a useful project. It does need someone to own the task, information the team may use, time for review and a way to handle failure. An AI readiness assessment should expose those conditions and turn gaps into specific work. The OECD's 2026 D4SME survey identifies constraints including time, skills and maintenance costs. Its sample is not representative of all SMEs or Mexican businesses. Use the findings as prompts for questions, not as a probability that your project will succeed.

Assess one proposed use case

Readiness varies by task. The same company may be ready for reviewed meeting summaries and unready for an assistant that accesses customer records. Start with a single problem brief, representative inputs and a process owner. Rate each condition ready, preparation needed or blocking.

Do not average away a blocking condition

Six ready answers do not cancel one unknown authorization decision. Likewise, an available technical partner does not replace a business owner. Mark the gap, the evidence needed and the person who will decide. A preparation task should have a concrete finish condition, such as confirming the current policy list or allocating review time. For document-level preparation, use the company knowledge preparation guide. This assessment stays focused on whether the organization can do that preparation and maintain it after a pilot.

A hypothetical readiness decision

Suppose a twelve-person services business wants an internal policy assistant. Assume its administrator owns the manuals, employees ask recurring questions and a technical partner is available. But the administrator also handles payroll and has no time reserved for answer checking. Some personnel files share a folder with general procedures. The business is not blocked from all AI work. It is unready for the proposed assistant drawing on that mixed folder. The sponsor narrows the source set to general procedures, excludes personnel files and assigns a supervisor two planned review sessions. The technical partner verifies access boundaries before using the material. The administrator confirms which procedures are current. The next decision depends on those completed tasks, not a readiness percentage. Once the conditions are met, a small offline test can compare answers with agreed references. If nobody can protect review time, postpone the pilot instead of assigning responsibility nominally.

Make readiness a recurring conversation

Review the checklist when the task gains users, sensitive information or integrations. Record changes to sources, owners and review workload. The NIST generative AI profile supports evaluating systems throughout their lifecycle; readiness can change after launch. Your output should be a brief preparation list with accountable owners, evidence and a review point. For help assessing one realistic use case, bring that list to a focused AI readiness assessment.