The AI Readiness Assessment Every SMB Should Run First

Every consultant has a maturity model. Most of them are nonsense. Five-stage transformation journeys with quadrants that take three months to fill in and produce a report nobody acts on.

The AI readiness assessment that actually matters for an SMB is much shorter. Five questions. If you can answer yes to all five, you're ready to start. If you can answer yes to three, you can probably start with help. If you can't get to three, fix the gaps first or the AI project will fail in ways that have nothing to do with AI.

This post walks the five questions, what a yes looks like, and what to do when the answer is no.

Why AI readiness is mostly not about technology

The framing trap most SMBs fall into is treating AI readiness as a technology question. Do we have a data warehouse? Are our systems on the cloud? Do we have an MLOps platform? None of these matter for the first AI project in an SMB.

What matters is whether your business can take a workflow, scope it tightly, hand it to a small project team, and operate the result for the next year. That's an organizational question, not a technology question. The SMBs that struggle with AI readiness usually struggle with the same things on any operational project. The SMBs that ship operations projects well will ship AI well.

I cover the organizational failure modes in detail elsewhere. This post is the diagnostic version.

Question 1: Do you have a specific workflow with a measurable hours-saved estimate?

Yes looks like: "Our SDR team spends about 14 hours a week triaging inbound leads, and I think an agent could absorb 8 of those hours. The team owns 200 leads a week."

No looks like: "We want to do something with AI in sales." Or "AI could help customer service." Or "Our team is overwhelmed and AI seems like it could help somewhere."

The bar here is specificity. A workflow has a name, a frequency, an owner, and an hours estimate. Without all four, the project will spend months trying to figure out what it's even building. Anyone who skips this step is funding a discovery exercise, not a build.

If you can't get to a specific workflow in a day or two of looking, the right move is a short discovery sprint. Pick the three departments where you suspect AI could help. Sit with one person in each for an hour and watch them work. Write down the repetitive tasks. By the end of the week you'll have a list of 30 to 60 candidate workflows, and the top five will be obvious.

Question 2: Is the data the workflow needs reachable and consistent?

Yes looks like: "The inbound leads come through our HubSpot form. We have an export. The fields are filled in for 90% of submissions. There's a clean CSV available."

No looks like: "The data is in three systems and nobody knows which one is right. We'd have to reconcile first."

The bar here is reachability, not perfection. Modern LLMs are surprisingly tolerant of messy data; they don't need a star schema. They do need the data to be in a place you can read from, in a format that's roughly consistent, with the key fields populated most of the time. If your CRM is half-empty because the sales team doesn't fill it in, the agent will inherit that gap.

The trap is waiting for a data warehouse to exist before doing AI. Most SMBs don't need a warehouse for their first three or four agents. They need an export from one system, sometimes joined with an export from a second. That's it.

A useful rule of thumb: if you could give an external analyst a Wednesday afternoon to put together the dataset, you have enough data to ship the agent. If they'd need a month, the data part is also a project.

Question 3: Is there a named operating owner with capacity?

Yes looks like: "Maria runs the support team. She has agreed to spend five hours a week for the first quarter on operating the agent, and one hour a week ongoing. Her manager has signed off on the time budget."

No looks like: "We'll figure out the owner at the handoff." Or "The vendor will support us."

This is the gap that kills the most SMB AI projects. The owner isn't optional. AI agents drift. The model behind them gets retrained by the vendor and the outputs shift. The world they're observing changes. The team's expectations shift. Without a named human watching the outputs and tuning the prompts, the agent goes stale fast.

In small companies, the right owner is usually the person who runs the workflow the agent is augmenting. Not the CTO. Not an external consultant. The line manager. They have the context to know when the agent is right and when it's wrong, and they have the credibility to make calls about what to do when the team disagrees with the agent.

If no one in the org has the capacity, that's a real readiness gap. Either restructure to free someone up, or accept that you can't take on an AI build right now and try again next quarter. The worst move is to pretend the owner question is solved and hope.

Question 4: Can your team make decisions inside two weeks?

Yes looks like: "We had a stand-up Monday, picked the build, approved the scope by Wednesday, and started Friday."

No looks like: "We need to bring this to the executive committee, which meets monthly."

The decision tempo question is sneaky. SMBs that take six weeks to approve a scope change will burn through their AI build budget on waiting. Models change. The original vendor recommendation might not be optimal by the time you've decided. The workflow itself might have shifted while the approval was in flight.

The bar is whether your team can make a build decision, a scope change, or a kill decision inside two weeks. Not next quarter. Not at the next board meeting. Two weeks.

If decisions take longer than this in your org, you can still do an AI project, but it has to be set up to need no decisions inside the build window. That means more pre-build scoping, a more conservative scope, and an explicit "we will not change scope between week 1 and week 8" agreement.

Question 5: Has your team finished an operational project in the last 12 months?

Yes looks like: "We replaced our ticketing system last fall. Three-month project, came in close to scope and on time."

No looks like: "The CRM migration has been ongoing for 18 months." Or "We tried to launch a new product line and it stalled."

Past behavior predicts future behavior. SMBs that have a recent track record of finishing operational projects will likely finish their AI project. SMBs whose recent track record is stalled initiatives will likely stall this one too.

This is not an AI problem. It's an execution problem. AI doesn't fix it. If you score no on question 5, the right move is usually to ship one non-AI operational project first to rebuild the execution muscle, then come back to AI when the team has a recent win.

This is also the question that's hardest for owners to answer honestly. Everyone wants to believe their team can finish projects. The accurate test is whether the projects that started 12 months ago have actually shipped, not whether they're still scheduled to ship.

How to use the scoring

Five yes answers: you're ready. Pick a first project and ship it.

Four yes answers: you can probably start with help. The gap on the one no should be addressed in the kickoff. Common pattern is no on operating ownership; the fix is to identify and free up the owner before signing the build contract.

Three yes answers: you can start with help, but the gaps will cost you. Plan for the build to take longer and cost more than the optimistic version. Be honest with your partner about which gaps exist.

Two or fewer yes answers: fix the gaps before starting an AI build. The build will fail in ways that look like AI problems but are actually the gaps you ignored. Save yourself the budget.

I cover why AI projects fail in more depth, and the failure modes there map directly to which questions an org said yes to without actually meaning it. The most common failure is saying yes to question 3 (owner) when the real answer is "we'll figure it out".

What this assessment is not

It's not an excuse to delay forever. SMBs that read this and use it to defer AI for two years are making a different mistake than the SMBs that rush in without doing the diagnostic. The point of the assessment is to tell you whether to start now, start in three months after fixing two specific gaps, or wait six months for the right moment.

It's also not the same as a roadmap. The readiness assessment tells you if you can start. The AI roadmap tells you what to do once you've started. Don't confuse them. Some owners run the assessment, decide they're ready, and then ask the same consultant to "write us the AI strategy". That's how you end up with a slide deck and no agent.

The honest version of the right next move, if you scored four or five yes answers, is to pick a workflow and ship in 90 days. The assessment was the warmup. The build is the work.