Most AI projects at small and mid-sized businesses do not fail because the model was not smart enough. They fail because the business was not ready to feed it. The team pours energy into picking the right AI tool and then discovers their data lives in three incompatible spreadsheets and a WhatsApp group. The model was never the bottleneck.

An honest ai readiness assessment measures the things that actually break projects, and almost none of them involve the AI itself. They involve your data, your processes, and whether anyone on your team is genuinely willing to change how they work. That is where I always start before recommending a single tool.

Below is a five-question self-assessment you can run in an afternoon. Score yourself honestly, because the point is not to feel ready. The point is to find the thing that would sink the project and fix it first.

The Five Questions

Give yourself a score from 0 to 2 on each. Zero means "not at all," one means "partially," two means "solidly yes."

  1. Can you access your data cleanly? Is the data the AI would need available in a consistent, exportable form, or is it scattered across paper, chat threads, and personal spreadsheets?
  2. Is the process you want to improve actually defined? Can you describe the current workflow step by step, or does everyone do it a little differently in their head?
  3. Do you have one measurable use case? Can you name a single task, with a number attached, that success would move? "Cut invoice entry time in half" is a use case. "Use AI" is not.
  4. Is there a willing person on the team? Is there at least one human who wants this to work and will adjust their routine, rather than quietly resisting it?
  5. Do you have a budget for iteration? AI projects are not one-and-done installs. Have you set aside money and time to tune the thing after the first version disappoints?

Add it up. Ten is rare. Most SMEs I assess land somewhere between four and seven, and that is fine. The score tells you where to work, not whether to quit.

Where Most Businesses Score Low

In practice, the low scores cluster on the first two questions: data access and process clarity. This is the unglamorous heart of ai readiness, and it is exactly the part the marketing skips.

A manufacturing supplier I worked with wanted AI to predict which customers were about to churn. Reasonable goal. But their sales history lived in one person's Excel file, updated by hand, with customer names spelled three different ways. No model on earth would help until that data was cleaned and centralized. The "AI project" was actually a data project wearing a fancier hat.

If you score low on data and process, that is your real starting point:

  • Low data score: get your core records into one consistent, exportable place before anything else. A clean database or even a well-structured shared sheet beats a genius model fed garbage.
  • Low process score: write the current workflow down, step by step, exactly as it happens today. You cannot automate a process you cannot describe.

The Use Case Question Is a Filter

Question three, the single measurable use case, quietly filters out most doomed projects. When a business cannot name one specific task with a number attached, they are not ready to build. They are ready to explore, which is a different and cheaper activity.

I push clients to finish this sentence: "Success means [this specific task] gets [faster / cheaper / more accurate] by [this amount] within [this timeframe]." If they cannot fill in the blanks, we do not start a build. We run a small experiment first.

This discipline is the same one that keeps any technology investment honest. If you have never measured whether a past tech project actually paid off, it is worth reading Measuring Digital Transformation ROI Without Fooling Yourself before you commit budget to an AI one.

Readiness Is Fixable, Not Fixed

The good news in every ai readiness assessment is that a low score is a to-do list, not a verdict. None of these five gaps are permanent. Clean data, define the process, pick one measurable use case, recruit a willing champion, set aside an iteration budget, and you have moved from "not ready" to "ready" in a matter of weeks, usually without spending much.

What you should not do is skip the assessment and buy the tool anyway. That is how businesses end up with an expensive AI subscription nobody uses and a quiet conclusion that "AI does not work for us." AI worked fine. The readiness did not exist yet.

If you want context on which AI investments actually mattered for businesses recently, and which were noise, The AI Year in Review: What Actually Mattered for Business is a useful reality check alongside this assessment.

The Practical Takeaway

Before you shop for AI tools, score your readiness honestly on the five questions.

  • If data access or process clarity is low, fix that first. It is almost always the real project.
  • Refuse to start a build until you have one measurable use case with a number attached.
  • Make sure a real person wants this and that you have budget to iterate past version one.
  • Treat a low score as a short to-do list, not a reason to give up.

Getting ready is cheaper than failing loudly, and it is most of the battle. If you want an outside read on where your business actually stands and what to fix first, that assessment is something I run with clients as a technology partner.