Back to Blog
    AI & Technology

    AI Readiness Assessment for a Small Business (2026): The 20-Question Scorecard

    The AI readiness assessment we run on a first call with a 5 to 50 person company: 20 questions across data, workflows, people, tools and risk, scored out of 100, with what each score should buy and real prices. Free scorecard included.

    Oct 27, 2026Updated Oct 31, 202612 min read

    TL;DR

    • An AI readiness assessment for a 5 to 50 person company comes down to five questions about how the business runs.
    • The first three: whether a computer can see your customers and jobs, whether the work is repeatable, and whether anyone will use what gets built.
    • The other two: whether your tools can talk to each other, and who checks what the machine says before a customer reads it.
    • Our free 20-question scorecard scores those five areas out of 100 in your browser, and each band maps to one purchase.
    • A low score is the norm rather than a verdict.

    Every AI readiness assessment ranking on Google is a consultancy's intake form. You answer thirty questions, a salesperson calls, and the answer is a six-figure roadmap. That is the right product for a hospital system. It is the wrong product for a plumbing company with eleven people, a Jobber account and a group text that runs dispatch. This guide is the assessment we run on the first call with a contractor or a professional firm, with the scoring shown and the prices attached to each result, so you can run it on yourself before anyone calls you.

    What readiness means when you have eleven employees

    The enterprise version measures model governance, data lakes and a chief AI officer. None of that is on the table under 50 people. Pretending it is produces the two failure modes we see most. One is the owner who buys an AI phone agent before the leads are logged anywhere, and the other is the owner who hires a consultant to write a strategy for a company whose quoting lives on carbon paper. Readiness at this size is whether the next automation will have anything to automate.

    The numbers say most small companies are early. The Census Bureau's Business Trends and Outlook Survey, published May 26, 2026, found AI use at 17 to 20 percent of US businesses. Thirty-seven percent of firms with 250 or more employees use it. Fewer than 20 percent of firms with four or fewer employees do. One caveat: the survey question changed in November 2025, so comparisons with earlier years are loose. Two things follow. Most of your competitors are not using AI either, so there is no urgency premium. And what separates a four-person shop from a 250-person company is that the bigger one already has its data in one place.

    The five areas, and why each one is a gate

    Five gates, in order 12345 DataWorkflowsPeopleToolsRisk Can a computersee the business? Is the workrepeatable? Will anyoneuse it? Can the softwaretalk to anything? Who checks whatthe machine says? 20 points each. A low score on the left makes everything to its right more expensive.
    The order matters. A company with a 4 on data and a 16 on tools has expensive software pointed at nothing.

    1. Data: can a computer see your business?

    The four questions here: where a new lead's name ends up, where quotes, jobs and invoices live, how long it takes to pull last month's leads, booked jobs and revenue, and whether inbound calls are logged anywhere. A company that scores under 8 here cannot automate follow-up, because there is nothing to follow up from, and cannot build a dashboard, because there is nothing to display. The fix is cheap and boring: one CRM every lead source feeds, a tracking number on the website and the Google listing, and the job software connected to QuickBooks. Our guides to CRM for contractors and which CRMs sync two ways with QuickBooks cover the choice. None of it is AI, and all of it is what the AI will read.

    2. Workflows: is the work repeatable?

    The questions: is the path from lead to booked job written down, and what happens when a quote goes quiet? How does a job get from the office to the crew, and can you name one task three or more people do the same way every time? Software can only repeat something that already repeats. If the quote follow-up depends on who remembers, an automated sequence will fire into a process nobody owns, and the first time it texts a customer who already paid, it gets switched off. The lead follow-up system is the written version of the first workflow most contractors need, and eight tasks you can automate today is the list to pick from once one path is written down.

    3. People: will anyone use it?

    The questions: who owns the software, and has anyone used ChatGPT, Claude or Gemini for work this month? How long does a new office hire take to learn the systems, and what happened the last time you changed a tool? This area kills more projects than the other four combined, and it is the one no vendor assesses, because the honest answer costs them the sale. A company where the owner is the only person who can change a setting is a company where every automation becomes the owner's problem at 9 pm. If the score here is low, the right first purchase is four sessions with the people who will run it, before any software. That is what the AI and Automation Coaching Sprint is for, at $1,000, and the one shared AI platform decision comes before any build.

    4. Tools: can your software talk to anything?

    The questions: can the CRM connect to other software, directly or through Zapier or Make, and does the phone system log calls and send texts? Do accounting and the job software sync, and how many logins does the office use before lunch? This is the area owners expect an assessment to be about, and it is the fourth gate, not the first. A CRM that connects to everything is worth nothing if the data in it is incomplete. The common pattern at a 60-point company is a good tool set wired to nothing: Jobber, QuickBooks, a VoIP line and a website that each hold a piece of the customer and never compare notes. The build that fixes that is usually the cheapest one on our list.

    5. Risk: who checks what the machine says?

    The questions: do you have a written AI use policy, and who sees an automated message before a customer does? Where does customer data live and who can reach it, and how would you find out a tool broke at 2 am? Small companies skip this area because it sounds like compliance. It is operations. We have watched a missed-call text back fire on 200 callers who had just spoken to the owner because one setting was wrong, and nobody knew for 17 days because nothing was watching. A policy at this size is one page: what the team may paste into an AI tool, what must never leave the CRM, and who approves anything a customer reads that a person did not write. That page is the AI policy template, a free Word download.

    Score it: the 20 questions

    Each question scores 0, 2, 3 or 4, and 5 for the best answer. The scorecard tool does the arithmetic live and shows the two weakest areas with the first fix in each. The table is the same twenty questions with the 0-point and 5-point answers, so you can score it on paper in a truck.

    AreaQuestion0 points5 points
    DataWhere does a new lead's name and number end up?Someone's head or a text threadA CRM, every lead, every source
    DataQuotes, jobs and invoices live inPaper or a phone's photo rollOne system, synced to accounting
    DataTime to pull last month's leads, jobs and revenueDays, and someone has to build itTen minutes, one screen
    DataInbound calls areNot logged anywhereLogged, recorded, tied to the customer
    WorkflowsIs lead to booked job written down?No, it depends who picks upWritten and followed
    WorkflowsA quote goes quiet. What happens?Nothing, unless they call backAn automatic sequence with a human at the end
    WorkflowsHow does a job reach the crew?A group text or a callDispatch software, no side channel
    WorkflowsOne task three people do the same wayCannot think of oneMost of the office runs that way
    PeopleWho owns the software?NobodyA named person with time for it
    PeopleAnyone used ChatGPT, Claude or Gemini for work this month?NoMost of the office, shared account
    PeopleA new office hire learns the systems inMonths, by shadowingDays, from written steps
    PeopleLast time you changed a tool, the teamKept the old wayAsked for it
    ToolsYour CRM or field service softwareDoes not existHas an API you already use
    ToolsYour phone systemIs a cell phoneLogs, records, texts, talks to the CRM
    ToolsAccounting and job softwareRe-keyed by handSync both ways
    ToolsLogins before lunchEight or more, some sharedOne or two, one login for everything
    RiskWritten AI use policy?NoWritten, signed, reviewed this year
    RiskWho sees an automated message before a customer?Nobody, it just sendsA person approves anything not a template
    RiskCustomer data livesIn texts, photos, personal emailIn the CRM, with roles
    RiskA tool breaks at 2 am. You find outWhen a customer complainsThat night, from an alert
    TotalUnder 40: fix data. 40 to 69: one build. 70+: a program.

    Score yourself in four minutes

    The scorecard runs in your browser, no email needed. Leave one if you want Alex to read your answers and reply with what to buy first, or that you need nothing yet.

    Open the AI Readiness Scorecard

    What each score should buy

    This is the part the consultancy assessments leave out, because their answer is always the same. Ours changes with the number, and every price below is on our pricing page.

    Under 40: not ready, and that is fine

    Automating a process that lives in a group text automates the mess. The first month is a $100 Jobber or Housecall Pro seat if you have neither, a tracking number with missed call text back, and one CRM every lead goes into. That is a month of office work, not a project, and most companies can do it without us. If the phone side is the blocker, Essentials at $350 a month includes call tracking, the text back and the review funnel on a site we build. Do not buy an AI phone agent, a dashboard or a consultant at this score. There is nothing for them to read.

    40 to 69: ready for one build

    Enough is in place to automate one thing well, and not enough for three. Pick the leak that costs the most. For most trades it is quotes that go quiet, which is a follow-up sequence; for an office that re-keys invoices, it is the QuickBooks sync; for a company missing calls, it is the AI phone agent. One fixed-price build from $1,000, scoped in writing, done in two to four weeks, and you see it on screen before it is built. A company at 55 that buys a 90-day program spends the first month doing the data work it could have done itself.

    70 and up: ready for a program

    The data is in one place, the work repeats, someone owns the software and the tools already connect. This is the company the Systems Sprint was built for: a custom CRM in month one, a proposal system in month two, an owner dashboard in month three. It is $2,500 a month for three months or $6,750 up front, and you own all of it. After that, $500 a month keeps it running and builds the next thing, or you run it yourself with the documentation. At this score the risk is scope rather than the build, which is why each month ships before the next one is paid for.

    Three things that fake readiness

    A ChatGPT subscription. The owner uses it daily, so the company "uses AI." The Census question is about business functions, and a chat window on one laptop does not qualify. Count it as a good sign for the People area and nothing more.

    An expensive CRM nobody fills in. ServiceTitan at $400 a month with half the leads in it scores lower on Data than a $29 Jobber seat with all of them. Every assessment that asks "which software do you use" instead of "where does a lead end up" gets this backwards.

    A consultant's roadmap. A 40-page document with phases and a maturity model, for a company that cannot pull last month's revenue in under a day. Strategy at this size is the order of operations above. If a proposal does not start with the customer list and the phone, it was written for a different company.

    What a 90-day plan looks like at each score

    The question we hear after the score is "then what, in order." Three plans, one per band, each a month per line.

    BandMonth 1Month 2Month 3
    Under 40Pick the CRM; move every lead source in; add trackingMissed call text back, review request, written lead-to-booked stepsConnect accounting; re-score. Most companies land in the 50s.
    40 to 69Scope one build in writing, see it on screenShip it, train the two people who touch itThirty days of fixes; measure the leak it was meant to close
    70 and upSystems map in week one, then the CRM buildProposal system, chased until paidOwner dashboard, training, documentation, hand over the keys

    For the under-40 band, the tracking number goes on both the site and the listing, and the phone gets fixed alongside the data. Above 70, the three months add up to a 90-day program.

    Want the score read by a person?

    Thirty minutes. We tell you whether you need a build, a program, or a software seat you can buy yourself this afternoon. If it is the seat, we say so.

    Book a strategy call

    Frequently asked questions

    What is an AI readiness assessment?

    A structured check of whether a business can get value from automation or AI before it spends on either. For a company under 50 people the useful version measures five things. Is the customer and job data in one system, is the work repeatable, and will anyone own the result? Can the tools connect, and does someone check what automated systems send to customers?

    What is a good AI readiness score for a small business?

    Seventy or above out of 100 on our scorecard means a 90-day program will stick. Forty to 69 means one well-chosen build will pay off. Under 40 means the customer list and the phone come first. Given that fewer than 20 percent of firms with four or fewer employees use AI at all, by the Census Bureau's count, most honest scores land under 60.

    How is this different from an AI audit?

    An AI audit reviews systems you already run: what the tools do, what they cost, where they fail. A readiness assessment runs before anything is bought. If you already have automations in place and want them checked, that is a different call, and the scorecard's Risk area is the place it overlaps.

    Do I need an AI strategy before I automate anything?

    Not a document. For a company this size the strategy is the order of operations: data in one place, then the phone, then one build, then a program. The score tells you which step you are on, and the plan table above is the whole roadmap.

    How much does AI readiness consulting cost?

    Enterprise firms charge $10,000 and up for an assessment and roadmap. Ours is free in the scorecard and 30 minutes on a call. The paid options after it are a $1,000 Coaching Sprint of four live sessions, one build from $1,000, or the Systems Sprint at $2,500 a month for three months.

    Does the scorecard send my answers anywhere?

    No. The score is computed in your browser. If you leave an email, your twenty answers go to Alex and you get one reply, not a sequence.

    AS

    Written by Alex Storey

    Founder of Skill Mammoth Digital. Builds the sites, runs the SEO and writes every article here. More about Alex.

    Book a Free Strategy Call