- In our experience, most businesses are more ready to start with AI than they think. What makes it stick is process design, security, and people approving what matters.
- The goal isn’t a tool for every task. It’s an AI operating system for your business, built as a path of steps that each pay off on their own.
- Start with one problem, even a big one like “our project management is a mess,” and climb from there.
- Your data doesn’t have to move or be in the cloud. It has to be reachable, including what’s in people’s heads, through meeting transcripts and recorded walkthroughs.
- Score yourself in about five minutes to see where you are on the path and your next step.
Every Friday afternoon, the office manager at Ferncliff Contracting, a 45-person commercial builder, used to chase timesheets. She’d text the crews, decode photos of crumpled paper and retype it all into payroll. It ate about four hours of her week (estimate).
Ferncliff is a made-up company, but we’ll follow it through this whole article.
Her first try was a chatbot. It worked for a week, then it got a pay rate wrong, nobody caught it, and the crew foreman lost trust in it. Sound familiar?
The fix wasn’t a better tool. It was the work around the tool: mapping how timesheets really flow, deciding where a person checks the output, and locking down what payroll data the AI can see. Now crews send a voice note from the truck, and the timesheets are waiting for her to approve.
That was Ferncliff’s first step, not its last. The owner’s real headache is bigger: project management. Updates rebuilt by hand every Friday, dates slipping without warning, questions lost in email. That’s where Ferncliff is climbing next, one step at a time.
How many businesses are actually using AI?
Plenty have started, but few have made it stick. Among small employers (1 to 499 employees), 46% say the business or its staff use AI, yet only 7% of those users have fully integrated it into how they work (Federal Reserve Small Business Credit Survey, March 2026).
Size plays a part. In the period ending May 3, 2026, 37% of US firms with 250 or more employees used AI in a business function, against under 20% of firms with four or fewer employees (US Census Bureau, May 2026; Census now labels this question “super intelligence (SI)”: see the BTOS page).
We don’t read that as big firms being smarter. We think they’re more likely to have people whose job includes the work around the tool, and a smaller firm can borrow that experience.
So the hard part isn’t starting. It’s making AI stick, then building on it. Below: five signs you’re ready, three common snags and their fixes, the path from a first win to a business that runs on AI, and the 16-question rubric.
How do I know if my business is ready for AI?
You’re probably more ready than you think. If you have one problem people complain about, work that repeats every week, information you can get to, managers open to trying something new and a way to tell whether it worked, you can start. Under each sign is where it tends to go wrong without someone who’s done it before.
1. Is there one problem everyone complains about or is costing you money?
You’re more ready than you think if you can name one problem that costs money, and the person who’d own fixing it. One problem with one owner beats a list of twenty ideas.
It doesn’t have to be small. Ask anyone at Ferncliff what’s broken and they’ll say “our project management is a mess.” That’s a dozen tasks tangled together. You don’t need a tool for each one. You need a first step that makes the whole thing better, and a path for the steps after it. Ferncliff’s senior project manager has offered to own it.
RAND’s interviews with AI practitioners found projects often fail because nobody pinned down the problem or how to measure success (RAND, 2024). A named problem with an owner settles both.
Where it goes wrong without expertise: naming the problem is easy. Mapping how the work really flows (who hands what to whom, where the exceptions hide) is where first tries often fall over.
You’re there if: you can say the problem and the owner’s name in one sentence. Scorecard question 5. Ready answer (3 points): “We’ve picked one, and a named person owns trying it.”
2. Does the same problem come round every week?
You’re more ready than you think if several people do the same work the same way every week. Volume turns a small saving into real hours back.
At Ferncliff, five project managers each spend Friday afternoon turning field notes, emails and schedule changes into an update for the building owner.
Generative AI users report saving about 2.2 hours a week, or 5.4% of work hours (St. Louis Fed, 2025, self-reported), so start where the hours pile up. If the steps aren’t written down, writing them is part of the first project, not a test to pass first.
Where it goes wrong without expertise: repeated work tempts people to let AI run it end to end. Setups that last decide where a person checks the output: every time at first, then by exception once it has earned trust. Skip that, and one bad output, like Ferncliff’s wrong pay rate, undoes the goodwill.
You’re there if: you can name work several people do every week the same way, and a new hire could follow its written steps. Scorecard questions 6 and 7. Ready answers: question 6 “Yes, kept current, and new staff learn from them.” (3 points); question 7 “Several people spend part of every day on it.” (2 points) or “Whole roles or teams are built around it.” (3 points).
3. Can you get to the information the work needs?
You’re more ready than you think if you can get to the information your first problem needs, wherever it lives. It doesn’t have to sit in one system or in the cloud. It has to be reachable.
At Ferncliff, job information is everywhere: drawings and schedules on a server in the office, budgets in each project manager’s spreadsheet, decisions buried in email, and the real sequencing know-how in the superintendent’s head. None of it has to move first. AI can work across a local server, a shared drive such as SharePoint, Google Drive or Dropbox, spreadsheets and inboxes, once it’s given access.
What’s in people’s heads can be reached too. Transcribe the weekly site meeting, and decisions stop living in one memory. Have the superintendent record a ten-minute walkthrough of how he sequences a job, and the team, and the AI, can draw on it. Tell people when a meeting is recorded, and check your obligations first.
Accuracy matters more than location. If leaders fix the main report by hand, list its three most common errors and fix them where the data is entered: a handful of small fixes, not a data project.
Where it goes wrong without expertise: the question isn’t only “can AI reach it?” but “what should it never see?” That’s a security decision, and the one do-it-yourself tries often skip. With payroll and employee records, like Ferncliff’s pay rates, it’s also a privacy call to make with your advisers before anything is built.
You’re there if: you can say where the information for your first problem lives and who can grant access to it, and nobody argues about the numbers in its main report. Scorecard questions 2 and 4. Ready answers (3 points): “We trust and act on them; errors are rare” and “Yes. Each main data area has a named owner”.
4. Are your managers open to trying it?
You’re more ready than you think if one person will own AI day to day and your managers would tell their teams it’s welcome on real work.
At Ferncliff, the operations manager has three hours a week set aside for AI, and each Monday meeting opens with someone showing one thing that worked on a real job.
In Gallup’s 2025 survey of more than 19,000 US workers, employees at organizations already investing in AI who strongly agreed their manager actively supports AI use were 2.1 times as likely to use it a few times a week or more (Gallup, November 2025): a link, not proof of cause.
Where it goes wrong without expertise: goodwill fades fast if the first step doesn’t fit how people already work. Rollout is a skill of its own: training people on their own tasks, showing them what changed in their week, and adjusting when they push back.
You’re there if: you can name the person who owns AI and point to a manager who has told their team to use it. Scorecard questions 9 and 11. Ready answers (3 points): “A named person with set time for it each week” and “Most encourage it and use it themselves”.
5. Could you tell whether it worked?
You’re more ready than you think if you know roughly how long the work takes today and can set aside a small, time-boxed amount to try something better.
Before changing anything, Ferncliff spent two weeks timing its Friday owner updates, set a small, fixed budget for a 60-day trial and put the comparison date in the calendar.
One owner we met this year was unhappy with what he was paying for an AI tool compared with what he got out of it. A starting number would have turned that feeling into a decision: keep it, change it, or stop.
Feelings are a poor guide. In a July 2025 randomized study, 16 experienced developers expected AI to make them 24% faster, yet they were 19% slower (METR, July 2025; METR now marks the study as out of date). The lesson travels: measure, don’t guess.
Where it goes wrong without expertise: it’s easy to measure the wrong thing, or nothing. Pick one number, capture it before you start, and check it at the end. That habit turns each step into the business case for the next.
You’re there if: you can say what the trial may spend, when it ends, and today’s number for the work. Scorecard questions 15, 16 and 8. Ready answers (3 points): “A set amount for the next 12 months, with someone responsible.”, “Today’s number and a target, checked before spending more.” and “We measure it regularly, such as time per job.”
What gets in the way, and how do you fix it?
We see three snags again and again in first AI projects: know-how that’s hard to get to, AI chosen before the problem, and staff using AI with no governance. Each first move is small: make one problem’s information reachable, name the problem before the tool, and put basic governance in place, each a few staff hours a week this month (estimate). None is a reason to hold off.
Snag 1: What if our know-how is scattered across spreadsheets, inboxes and people’s heads?
Start by making the information for one problem reachable, without moving it. AI can only work from what it can get to, and where copies conflict, it will pick one and sound sure about it.
As one of our team puts it, once you get what’s in people’s heads out of their heads, you can start to do something with AI.
At Ferncliff, pricing know-how sits with two senior estimators, and job history lives in old email and closed-out folders on the office server. That becomes usable once the estimators’ bid reviews are transcribed and the folders are opened to AI.
Easy first move: List where the information for your first problem lives: server folders, spreadsheets, inboxes and people. Give access to what it needs, mark the current version of anything with several copies, then transcribe one recurring meeting and record two short walkthroughs. Owner: operations or office manager. Time: roughly 12–16 staff hours over the month (estimate).
Where expertise helps: deciding what AI may read and what stays out, so reaching more never means exposing more.
Snag 2: What if we want AI but can’t name the problem?
Start by naming the problem, so AI has a job to do. A tool bought before the problem is named often ends up unused; buy one per task and you get a pile of tools that don’t share information.
The projects that work start the other way round.
At Ferncliff, the owner’s first instinct was “get us an AI tool.” Three vendors pitched, each selling a fix for one task: meeting notes, scheduling, RFIs. Only when the team put those tasks side by side did the owner see them as one problem, project management, with one person to own it.
Easy first move: List your five most repeated tasks with rough hours per week for each, then group them under the bigger problem they belong to. Circle the biggest problem and name one person to own the first 30-day step. Owner: the executive asked to “look into AI” (often the COO or an operations head). Time: about 3–4 staff hours, including one half-hour team conversation (estimate).
Where expertise helps: turning “the problem” into a measurable target and a sequence of steps.
Snag 3: What if people are already using AI with no governance?
Start with written guidance, then build governance into how your AI is set up.
In ISACA’s 2025 poll of more than 3,000 professionals, only 28% said their organization had a formal, comprehensive AI policy (ISACA, July 2025).
At Ferncliff, estimators paste bids, with client names, pricing and margins, into free chatbots to tidy up proposals. Guidance tells them to stop.
Easy first move: Write one page of AI guidance: approved tools, who approves new ones, and a five-line “never paste” list. Move the most-used AI tool onto company accounts, check its data settings, and start a simple list of each AI tool, what it can reach and who owns it. Owner: operations lead or IT manager, signed off by the CEO. Time: about 8–10 staff hours over the month, plus a 15-minute team briefing (estimate).
Where expertise helps: building in the right controls.
None of the three fixes is about a smarter AI tool. They’re about how work flows, what AI can reach, and how it’s governed, the part a do-it-yourself try leaves out. They’re also the first rungs of the path.
What the path looks like
The goal isn’t a tool for every task. It’s an AI operating system for your business: one connected way of working where AI takes on more of the routine work over time, people approve what matters, and the process gets a little better at every step. You climb it, and each step pays off on its own.
What changes as you climb isn’t how capable the AI is; today’s assistants already connect to email, files and apps and can act. What grows is how far you can trust it: company-controlled access, answers grounded in your own records with sources shown, approval before it acts, testing before changes go live, a log with a named owner, and an off switch.
Here’s Ferncliff’s “our project management is a mess” as four steps, one per scorecard level.
Step 1 · Getting started
Capture the work
- AI does
- Turns site and owner meetings into notes and action items automatically.
- People approve
- The project manager checks the notes before they go out.
- Gets better
- Decisions stop getting lost, and nobody retypes notes on Friday.
Step 2 · Building foundations
One view of every job
- AI does
- Pulls one weekly status view from the systems Ferncliff already uses: the schedule, the job-cost spreadsheets, the office server and email.
- People approve
- The operations manager reviews it before the Monday meeting.
- Gets better
- One agreed answer to "are we on track?", and the Friday rebuild is gone.
Step 3 · Ready to pilot
AI flags and drafts
- AI does
- Flags dates that are slipping and drafts the owner update and the follow-ups to subcontractors.
- People approve
- The project manager approves every message before it's sent.
- Gets better
- Problems surface sooner, and the team checks hours saved against its starting number.
Step 4 · Ready to scale
Agents coordinate the flow
- AI does
- Agents route RFIs, submittals and change orders, chase answers and keep the log current.
- People approve
- Everything runs inside approval steps; anything touching money, contracts or safety waits for a named person.
- Gets better
- The process runs the same way on every job, with a record of who approved what.
Steps 1 and 2 use tools companies run today. Steps 3 and 4 need integration work and steady human oversight, so the earlier steps come first: each builds the access, habits and trust the next needs.
Where are you today? Your scorecard level is your step on the path, and the path above shows the next one. Getting started? Capture the work on your biggest problem. Ready to pilot? Run a measured trial with people approving the output. Every level has a next step.
What should I do next?
Whatever you scored, the easiest next step is the same: a free call with Ninja Concepts. Bring the one problem you’d fix first, and we’ll show you where you are on the path, your next step, and the process, security and checkpoints that make it stick.
If you scored Getting started or Building foundations, we help you build the foundations: one problem, the information it needs and simple guidance.
If you scored Ready to pilot or Ready to scale, we help you move fast and scale: scope the next workflow, decide where people approve the output, and connect what works.
That’s how Ferncliff’s second try went. It didn’t start with a tool. It started with one painful Friday afternoon and someone who had mapped work like it before.
How do I score my business?
Answer the 16 questions below for your company as it is today, add up the points, and you’ll have a score out of 48 in about five minutes. It isn’t a pass-or-fail test: every level comes with a good first move, and the questions show you exactly where to start.
How scoring works
- 16 questions, each scored 0 to 3. “Not sure” scores 0.
- Your overall score is the total, out of 48. Every question counts the same, so data and process (4 questions each) count more than budget (2).
- Five areas: data (12 points), process (12), people (9), rules and governance (9), budget (6).
- Levels: Getting started 0–14 · Building foundations 15–26 · Ready to pilot 27–38 · Ready to scale 39–48.
- An area needs work when it scores below half its points. Your report names up to two, the lowest first.
- The areas you pick to focus on, and the optional budget-stage question, don’t change your score.
Score yourself in about 5 minutes. Answer each question for your company as it is today. Give yourself the points next to your answer. “Not sure” scores 0. Add up all 16 for a total out of 48.
Data (4 questions, 12 points)
1. When work needs your company’s business information, can people get to it?
Wherever it’s kept: business software, a local server, shared drives, spreadsheets, email, or recorded meetings and walkthroughs.
| Points | Answer |
|---|---|
| 0 | Rarely. Much of it is only in people’s heads |
| 1 | Sometimes, but much is scattered or never written down |
| 2 | Mostly, with a few gaps |
| 3 | Almost always, wherever it’s kept |
| 0 | Not sure |
2. When leaders look at your company’s key report, what usually happens next?
| Points | Answer |
|---|---|
| 0 | We usually argue about whether the numbers are right |
| 1 | We often fix the numbers by hand first |
| 2 | We trust most numbers, with a few known problem areas |
| 3 | We trust and act on them; errors are rare |
| 0 | Not sure |
3. Can information from different places be pulled together without retyping?
| Points | Answer |
|---|---|
| 0 | No. We retype or copy and paste it |
| 1 | Sometimes. We export and combine files now and then |
| 2 | For the most important information; the rest is manual |
| 3 | Yes. Key information comes together automatically |
| 0 | Not sure |
4. Is a named person responsible for keeping your company’s key data accurate?
| Points | Answer |
|---|---|
| 0 | No one specific. Everyone and no one |
| 1 | Informally, whoever cares most, or IT by default |
| 2 | Yes, for some areas, like finance for financial data |
| 3 | Yes. Each main data area has a named owner |
| 0 | Not sure |
Process (4 questions, 12 points)
5. Have you picked one problem where you want to try AI first?
| Points | Answer |
|---|---|
| 0 | Not yet. We’re still exploring what AI could do. |
| 1 | We have a list of ideas but haven’t picked one. |
| 2 | We’ve picked one, but no one is responsible for it. |
| 3 | We’ve picked one, and a named person owns trying it. |
| 0 | Not sure |
6. Are the steps for your team’s most repeated task written down?
| Points | Answer |
|---|---|
| 0 | No. It lives in one or two people’s heads. |
| 1 | Some notes exist, but they’re out of date. |
| 2 | Yes, written down, and most people follow them. |
| 3 | Yes, kept current, and new staff learn from them. |
| 0 | Not sure |
7. How much of your team’s week goes to work that follows the same pattern each time?
| Points | Answer |
|---|---|
| 0 | Very little of anyone’s week. |
| 1 | A few hours a week across the team. |
| 2 | Several people spend part of every day on it. |
| 3 | Whole roles or teams are built around it. |
| 0 | Not sure |
8. Do you know how long your most repeated task takes today, or what it costs?
| Points | Answer |
|---|---|
| 0 | No one has looked at it. |
| 1 | We have a rough guess. |
| 2 | We track it loosely, such as in a spreadsheet. |
| 3 | We measure it regularly, such as time per job. |
| 0 | Not sure |
People (3 questions, 9 points)
9. Apart from any one project, is one person responsible for AI in your company day to day?
| Points | Answer |
|---|---|
| 0 | Nobody has been asked to own it yet |
| 1 | Whoever is interested, in their spare time |
| 2 | A named person, fitted around their main job |
| 3 | A named person with set time for it each week |
| 0 | Not sure |
10. What hands-on AI practice have your people had, using their own day-to-day tasks?
| Points | Answer |
|---|---|
| 0 | None yet |
| 1 | People teach themselves, from videos or trial and error |
| 2 | Some teams have had a session on their own tasks |
| 3 | Most teams have practised on their own tasks, with follow-ups |
| 0 | Not sure |
11. How do your managers and team leads treat AI use on work tasks?
| Points | Answer |
|---|---|
| 0 | Most discourage it or don’t allow it |
| 1 | Most say nothing either way |
| 2 | Some encourage it on work tasks |
| 3 | Most encourage it and use it themselves |
| 0 | Not sure |
Rules and governance (3 questions, 9 points)
12. Which best describes how your company has told staff to use AI tools?
| Points | Answer |
|---|---|
| 0 | Nothing yet. People decide for themselves. |
| 1 | Leaders have mentioned it, but nothing’s written down. |
| 2 | Written guidance, but not everyone has it, or it’s dated. |
| 3 | Written guidance everyone has received, reviewed in the past year. |
| 0 | Not sure |
13. Have you told staff what information must never go into AI tools?
| Points | Answer |
|---|---|
| 0 | Not yet. It hasn’t come up. |
| 1 | A general reminder, like “don’t share anything confidential.” |
| 2 | A written list with specific examples. |
| 3 | A written list, and we’ve checked the data settings on our tools. |
| 0 | Not sure |
14. When a team wants to start using a new AI tool, what happens?
| Points | Answer |
|---|---|
| 0 | They just start using it. |
| 1 | They usually mention it to their manager first. |
| 2 | One named person or group approves new AI tools. |
| 3 | A named owner approves each one against a checklist and logs it. |
| 0 | Not sure |
Budget (2 questions, 6 points)
15. Is money set aside for AI over the next 12 months?
| Points | Answer |
|---|---|
| 0 | No, and it hasn’t come up. |
| 1 | We’ve talked about it, but nothing is set aside. |
| 2 | We pay for some AI tools, but no set amount. |
| 3 | A set amount for the next 12 months, with someone responsible. |
| 0 | Not sure |
16. Before spending more on AI, do you write down what result would justify it?
| Points | Answer |
|---|---|
| 0 | No. Gut feel or a demo decides. |
| 1 | A general goal, like “save time,” but no number. |
| 2 | A number for some projects, rarely checked after. |
| 3 | Today’s number and a target, checked before spending more. |
| 0 | Not sure |
Read your score
| Total (out of 48) | Level |
|---|---|
| 0–14 | Getting started |
| 15–26 | Building foundations |
| 27–38 | Ready to pilot |
| 39–48 | Ready to scale |
Find the areas that need work. An area needs work when it scores below half its points: data 5 or fewer of 12, process 5 or fewer of 12, people 4 or fewer of 9, rules and governance 4 or fewer of 9, budget 2 or fewer of 6. If more than two areas need work, start with the two lowest by percent. On a tie, take them in this order: process, data, people, governance, budget.
Your level is your step on the path. See “What the path looks like” above for what that step looks like and what comes next.
View more at ninjaconcepts.ai. This article and others like it: www.ninjaconcepts.ai/resources/is-my-business-ready-for-ai
