SMB Scaler
Guide24 questions · 6 areas

AI Readiness Assessment for Small Businesses

A free, self-scored AI readiness assessment for businesses with 5 to 500 people: 24 questions across six areas, a score, and what to fix before you spend anything on AI.

Process Management

Most AI readiness assessments are written for enterprises with a data team, a CIO and a transformation budget. They score things a ten- or fifty-person firm doesn't have and doesn't need, and they rarely answer the question an owner is actually asking: is there one workflow here that AI can take over, and are we set up to let it?

This assessment is built for that question. It looks at one business, not an industry, and at the work people do every day, not at the software you own. Answer 24 questions across six areas, add up the score, and you'll know whether to build now, run a tightly scoped pilot, or fix a few basics first, and which basics.

Who this assessment is for

  • Owners and operators of service businesses with roughly 5 to 500 people: accounting and bookkeeping firms, insurance agencies and MGAs, freight brokers and forwarders, property managers and other services firms.
  • Teams that have tried an AI tool or two and aren't sure why it didn't stick.
  • Anyone about to talk to an AI vendor or consultant who wants to walk in knowing where the business stands.

How to run it

  1. Pick one workflow before you start: the one that most often forces you to hire, wait or turn work away. The assessment is sharper scored against a real workflow than against the whole business.
  2. Score it with two people: the owner or manager, and the person who does the work every day. Where your answers differ, that difference is itself a finding.
  3. Answer each question Yes, Partly or No. Be strict: "we could find that out" is a No.
  4. Total the score, read your band, then go back to the lowest-scoring area. That's where your first fix is.

Scoring: Yes = 2, Partly = 1, No = 0. The highest possible score is 48. Pick your answers below and the score adds itself up, or note them on paper.

The assessment

A named bottleneck

01A named bottleneck

0 / 8

AI pays off when it removes the constraint that caps what the business can take on. Without a named bottleneck, projects drift toward whatever demo looked best.

  1. Can you name the one workflow that most often forces you to hire, make clients wait or turn work away?
  2. Do the owner and the people doing the work agree on what that workflow is?
  3. Can you say what the business would do with the time or capacity if that bottleneck went away (more clients, a new service line, faster turnaround)?
  4. Is that workflow important enough that fixing it would change this year's results, not just save a few hours?

If this scores low: Spend a week logging where time goes before choosing any tool. Ask everyone what they'd stop doing if they could, and look for the task that keeps coming up.

How the work is done

02How the work is done

0 / 8

AI can follow a process; it can't settle an argument about what the process is. Work that runs differently depending on who does it has to be pinned down first.

  1. Is the workflow written down step by step, even roughly (a checklist, a doc, a recorded walkthrough)?
  2. Do different people on the team do it mostly the same way?
  3. Can you list the exceptions (the cases that need a judgment call) and who handles them?
  4. Do you know where the handoffs are, between people, teams or systems, and where work stalls at them?

If this scores low: Map the workflow before automating it: sit with the person who does it, write down every step and every exception, and agree on one way of doing it.

Inputs and data

03Inputs and data

0 / 8

Every AI system runs on what you feed it. The question isn't whether you have "big data"; it's whether the documents and records this workflow uses can be reached and read.

  1. Do you know every place the inputs to this workflow come from (email, PDFs, portals, spreadsheets, a core system)?
  2. Can those inputs be reached digitally, without someone retyping them from paper or screenshots?
  3. Do the systems you rely on let you export data or connect to them (an export, an integration or an API)?
  4. Is there one place you'd trust as the source of truth for the records this workflow touches?

If this scores low: List every input and where it lives. Ask each software vendor what export or integration options your plan includes; the answer often decides what's possible.

Volume and repetition

04Volume and repetition

0 / 8

A task done three times a month rarely justifies a build. A task done thirty times a day, the same way, usually does.

  1. Does this workflow run often: daily or many times a week?
  2. Do most cases follow the same pattern, with judgment calls the exception rather than the rule?
  3. Is a meaningful share of someone's week spent on it?
  4. Would volume grow if you could handle it faster (more quotes, more clients, more files)?

If this scores low: Pick a different workflow. Low-volume, high-judgment work is a poor first project; look for the repetitive task that sits right before or after it.

People and ownership

05People and ownership

0 / 8

Most AI projects in small firms stall on people, not technology: nobody owns the system, the team doesn't trust it, or the expert who knows the work was never asked.

  1. Is there one named person who will own the system after it's built?
  2. Will the person who does this work today help design and test the new way?
  3. Can the team set aside a few hours a week for a month to test and correct it?
  4. Has the team been told what changes for them, and what happens to the time it frees up?

If this scores low: Name an owner and involve the person who does the work from day one. Say plainly that the goal is taking work off their plate, not taking their job.

Baseline and guardrails

06Baseline and guardrails

0 / 8

Without a before number you can't show an after. And in regulated or client-confidential work, you need to know what the system may and may not touch before it touches anything.

  1. Do you know roughly how long this workflow takes today (hours per week, or turnaround per item)?
  2. Do you track its error rate or rework, even informally?
  3. Do you know which client or regulated data can't leave your systems or be sent to an outside AI service?
  4. Is there a review step where a person checks the output before it reaches a client, carrier or regulator?

If this scores low: Time the workflow for two weeks before you change it. Check your client contracts and the rules that apply to you (your state regulator, carrier agreements or professional body) for what data can be shared.

Your score

0 / 48

0 of 24 questions answered. Unanswered questions count as zero.

What your score means

These bands are a rule of thumb for where to start, not a grade. A low score in one area matters more than the total: a business that scores well everywhere except data still has a data problem to solve first.

0–19 points

Not ready yet: fix the basics

The gaps are in the foundations, not the technology. Buying a tool now is likely to add work rather than remove it.

  • Start with your lowest-scoring area and its first fix above.
  • Write the workflow down and time it for two weeks.
  • Re-score in a month; most firms move up a band once the work is mapped.

20–35 points

Ready for a scoped pilot

You have a real bottleneck and most of what's needed to work on it. Keep the first project narrow: one workflow, one owner, one measure.

  • Close the gaps in any area that scored 3 or less before you build.
  • Set the success measure up front, using your baseline.
  • Plan for the pilot to become the real system, not a demo that's thrown away.

36–48 points

Ready to build

The workflow is clear, the inputs are reachable, the team is in and you can measure the result. The risk now is waiting, or picking a tool that doesn't fit how the work is done.

  • Build around the workflow as it runs, not around a generic tool's features.
  • Keep the person who does the work in the loop as reviewer for the first weeks.
  • Decide now what the freed-up capacity goes to, so it compounds instead of disappearing.

Common mistakes when assessing AI readiness

Scoring the software, not the work. Owning a CRM or a document system doesn't make a business ready. What matters is whether the work that runs through it is consistent, reachable and measured.

Treating a pilot as proof. A pilot shows a model can do a task once in good conditions. Readiness is whether your business can feed it the real inputs, every day, with someone owning the result.

Assessing the whole company at once. A company-wide score averages a ready workflow with an unready one and hides both. Score one workflow at a time.

Leaving out the person who does the work. Owners tend to score process and data higher than the people doing the work would. Score it together.

Frequently asked questions

What is an AI readiness assessment?

A structured check of whether a business is set up to put AI to work: whether it has a clear problem to solve, a defined process, reachable data, enough volume, people to own it and a way to measure the result. It tells you what to fix before you spend money, not which tool to buy.

How long does it take?

Answering the 24 questions takes under an hour for one workflow. Finding the true answers (timing the work, listing every input) can take a week or two, and that's usually where the value is.

Do we need clean data or a data team first?

No. A small business needs the inputs for one workflow to be reachable and reasonably consistent, not a company-wide data program. The data area of the assessment checks exactly that.

Is a free readiness assessment from a vendor enough?

It can be a useful start, but check what it scores. An assessment from a software vendor often measures readiness for that vendor's product. Make sure yours starts from your bottleneck and your workflow.

We scored low. Should we skip AI?

No: a low score tells you where to start. Mapping the workflow, naming an owner and taking a baseline are cheap, and they make any later project faster and more likely to pay back.

Go deeper