Search for an AI readiness assessment and you will find tools built for enterprises: long questionnaires, a gated PDF, and a sales call at the end. This one is built for a firm of 10 to 50 people, it fits on a page, and nothing is gated. Twelve questions in four groups, with a plain reading of what a “no” means. Answer them before you buy seats, not after.
The reason to bother is in the numbers. In Goldman Sachs' survey of 1,256 small business owners this past winter, 76 percent said they are already using AI, and 93 percent of those reported a positive impact. But only 14 percent said AI is “fully embedded in their core operations,” and 73 percent said they would benefit from more training and implementation support. Adoption is not the problem. Getting past the first few use cases is.
Part one: the work
1. Can you name three recurring tasks, by name, that take real hours every week? Not “admin” or “reporting,” but “the Friday client status email” and “re-keying invoices into the project tool.” A no here means the first month of any rollout should be a two-week time audit, not a tool.
2. For each of those tasks, does a finished example exist that everyone would agree is good? AI tools are very good at matching a pattern and poor at inventing your standards. If there is no agreed example, you will spend the pilot arguing about quality instead of measuring it.
3. Is there one workflow where a human already reviews the output before it goes anywhere? That workflow is your safest first project, because the review step already exists. If every process ends with something going straight to a client, add a review step to one of them before you begin.
Part two: the data and the permissions
4. If a new hire started Monday, would you know which folders, channels, and records they should and shouldn't see? This matters more than most owners expect, because Claude's connectors inherit each person's existing permissions from the source system. If your sharing is clean, Claude's access is clean. If it isn't, a connector will reflect that on day one. We covered this in detail in our guide to connectors for small teams.
5. Do you know where the sensitive material lives? Payroll, banking, HR files, client contracts with confidentiality terms. A yes means you can keep those folders out of scope from the start. A no means you don't yet know what you'd be exposing, and finding out should come first.
6. Do any client contracts say something about where data may be processed or by whom? Third-party services on the other end of a connector process data under their own terms, and Anthropic notes those may be located outside the United States. Agencies with enterprise clients often have language about this that nobody has read since signing.
Part three: the people
7. Is there one person, other than the owner, who is curious enough to try things and respected enough that others will follow? This is the single best predictor we see. Tools get adopted when a peer shows a peer, and stall when the owner demos something nobody else touches.
8. Have you told the team, in writing, what is and isn't allowed? Most small firms have people using AI already, on personal accounts, with company information. Goldman's respondents named data privacy as one of their three main barriers to going deeper. A short acceptable-use note does more for privacy than any setting, because it brings the existing use into the open.
9. Is there time in the calendar for it? Sixty-four percent of small businesses in a recent Business.com study said they are likely to launch AI training programs; far fewer have blocked the hours. An hour a week, for a month, with the tasks from question one in front of people, beats a half-day workshop nobody applies.
Part four: ownership and measurement
10. Who decides which tools get turned on, and who can say no? On a Team plan this is the Owner role, and it should be a person, not a committee. If the answer is “whoever asks first,” you will end up with a directory full of half-connected tools.
11. What number would tell you, at 30 days, whether it worked? Hours saved on the three tasks from question one is the usual answer, and it is a fine one. Pick it before you start. Deciding what to measure after the fact is how every pilot becomes a success in the retelling.
12. What happens to the workflow if the tool goes away? A yes means the process is documented and a person could pick it back up. A no is not disqualifying, but it tells you to write the workflow down as you automate it, which is a good habit regardless.
Reading your score
Count the yeses. Ten or more: buy the seats and start with the workflow from question three. Six to nine: spend two weeks on the noes in parts one and two first; they are cheap to fix and expensive to skip. Five or fewer: the right first project is not an AI project. It is the time audit, the folder cleanup, and the one-paragraph policy. Do those, and the tools will land on something solid.
If you want a second set of eyes on your answers, or someone to run the first 30 days with the team, that is the work Cedar & Loam does with agencies and small firms. Proudly based in Kansas City. Say hello.
Sources
- Small Businesses Embrace AI, But Need Training and Support to Fully Harness It, Goldman Sachs 10,000 Small Businesses Voices survey of 1,256 owners, January 27 to February 4, 2026
- 2026 Small Business AI Outlook Report, Business.com and Dialog, 1,009 respondents
- Use connectors to extend Claude's capabilities, Anthropic Help Center