An AI readiness checklist tells you whether a specific process is ready for AI before you spend money on it. Answer the 25 questions below for one process, count the yeses in each of the five areas (process, data, systems, people and risk), and you’ll know whether to start a pilot, do some groundwork first or pick a different process. Every "no" comes with what to fix.
How to use this checklist
Check the boxes as you go (they reset when you leave the page), then use the scoring table further down. If you’d rather work through it with someone who has seen a lot of these projects, that’s what our AI readiness assessment is for. For a quicker, broader read first, the interactive self-check below gives you a rough score across eight questions in a couple of minutes.
Interactive AI readiness self-check with eight multiple-choice questions about your data, processes, tools, team, volume of repetitive work and goals. Each answer scores from 0 to 3 and every area gets an indicative score from 0 to 10, the plain average of its answers, with a short explanation of what it means. It is a rough self-check, not a formal assessment, and it links to the contact form to talk through the results.
AI readiness self-check
Illustrative examplePick an answer to continue.
Plain-language notes for each area, based only on the answers you gave.
-
Data Early Developing Strong
Information is scattered or unreliable, so AI would have little to work from. Gathering one process’s data in one place is the first win.
The data exists but needs checking. An AI project would include a cleanup step for the records it relies on.
Your data is organized and trusted, which makes agents and automations much easier to build and test.
-
Processes Early Developing Strong
Steps live in people’s heads. Writing down one process end to end is worth doing before automating it.
Some steps are written down. Filling in the exceptions and hand-offs makes a process ready to automate.
Clear, documented steps are exactly what automation needs; you can point to where the time goes.
-
Tools Early Developing Strong
Manual copy-paste between tools is common. That is often the easiest place to start, but it may need a connector built.
Some tools connect. Expect a mix of ready-made integrations and a few custom links.
APIs and existing automations mean AI can plug into your tools without changing how people work.
-
Team Early Developing Strong
Without an owner and some buy-in, projects stall. Naming one person and starting small helps.
There is interest and someone to lead. Protecting a few hours a week for them keeps a pilot moving.
An owner with backing and a willing team give a pilot its best shot at sticking.
-
Volume Early Developing Strong
The task is infrequent, so the case for automating it is weaker. Look for the work that repeats most.
Enough repetition to matter. Automating it mostly frees attention rather than whole roles.
High-volume, repetitive work is where AI agents and automations tend to pay back fastest.
-
Goals Early Developing Strong
No problem picked yet. A short workshop to list and rank candidate processes is a good first step.
A general direction is a start. Turning it into one specific process with a clear success check is next.
A clear target and a way to judge success make it easy to scope, build and evaluate a first project.
Your indicative profile
Rough self-check · not a formal assessmentScores fill in as you answer. Each one is the average of that area’s answers, on a 0–10 scale.
-
Data Not answered yet Early Developing Strong /10
-
Processes Not answered yet Early Developing Strong /10
-
Tools Not answered yet Early Developing Strong /10
-
Team Not answered yet Early Developing Strong /10
-
Volume Not answered yet Early Developing Strong /10
-
Goals Not answered yet Early Developing Strong /10
Answer the questions to see where you stand.
Groundwork first. A short discovery to map one process and tidy its data usually comes before any build.
A solid base. A small, well-scoped pilot on one process is a sensible next step.
Well placed to pilot. The work is choosing the right first process and measuring it properly.
Indicative only: it reflects the eight answers you gave and nothing else. It is not a formal assessment, audit or guarantee; talking through your actual processes is what tells you what to build.
Share what you’re working on and we’ll reply with next steps.
That self-check gives you a feel for the business as a whole. The 25 questions that follow dig deeper into one process, which is what you need before you commit a budget.
1. Process readiness (questions 1–5)
AI follows rules. It can’t invent a process that doesn’t exist yet.
Process
If you answered no: document the process first. Sit down with the person who does it, write out each step and the five most common exceptions, and agree on a baseline (time per item, items per week). That baseline is also how you’ll measure success later.
2. Data readiness (questions 6–10)
The AI is only as good as what it reads.
Data
If you answered no: gather 50 to 200 real past examples with their correct outcomes. They become your test set. Clean up documents that contradict each other, and flag sensitive data so access rules can be built in from day one.
3. Systems readiness (questions 11–15)
Most AI projects are integration projects in disguise.
Systems
If you answered no: get admin access back, ask vendors about APIs and test environments, and name an owner for integrations. If a key system has no API at all, that shapes the whole design, and it’s better to find out now. Our guide on how to integrate AI into existing software covers the options.
4. People readiness (questions 16–20)
Technology rarely sinks AI projects. A lack of ownership does.
People
If you answered no: name an owner and a reviewer before anything gets built, and give the team a short, practical briefing. When skills are the gap, structured AI training for your team is often the fastest fix.
5. Risk readiness (questions 21–25)
This area decides how much autonomy the AI can safely have.
Risk
If you answered no: write a one-page AI usage policy, list the hard limits, and design the escalation paths before the build. Our guide to writing an AI governance policy has a template structure. If calls are involved, read our explainers on HIPAA and AI voice agents and the TCPA and AI phone calls.
How to read your AI readiness score
Count the yeses in each area (five at most per area).
| Area score | What it means | What to do |
|---|---|---|
| 5 | Ready in this area | Move on |
| 3–4 | Mostly ready; the gaps are specific | Close them as part of project setup |
| 0–2 | Not ready in this area | Fix it before the project starts |
Then look across all five:
- Every area at 3 or more: this process is a good pilot candidate. Use your answer to question 4 as the baseline for success and start small.
- Any area at 2 or less: do the groundwork first. It’s usually documentation, data cleanup or access, which takes far less time than recovering from a failed project.
- Risk at 3 or less: whatever the other scores say, deal with risk first.
An example: a regional insurance agency
This is an illustrative example, not a client. An independent insurance agency in Ohio wants AI to sort its shared service inbox: certificate requests, policy changes, billing questions and claims.
- Process: 5/5. Account managers can describe how each type of email is handled, and these emails come in all day.
- Data: 3/5. Procedures are written down, but some carrier rules live in one senior account manager’s notes, and there are no labeled examples yet.
- Systems: 2/5. The agency management system has an API, but nobody knows who holds admin access, and there’s no test environment.
- People: 4/5. A service manager will own it, and the team has been involved from the start.
- Risk: 3/5. Claims emails need same-day attention from a person, and there’s no AI usage policy.
Verdict: not ready to build, but close. Two weeks of groundwork (get admin access back, set up a test account, export a few hundred past emails with their correct categories, write a one-page policy and a rule that sends anything mentioning a claim straight to a person) would make this a strong first project.
What to do next
From checklist to pilot
-
Step 1
Score two or three processes
comparing them often turns up a better first candidate than the one you started with
-
Step 2
Close the gaps in the weakest area
usually documentation, access or a test set
-
Step 3
Define success in numbers
time per item, response time or error rate, measured against today’s baseline
-
Step 4
Run a narrow pilot
one process, real cases, people reviewing everything for the first few weeks
-
Step 5
Decide based on evidence
expand, adjust or stop depending on what the logs and the baseline show
For a wider look at where AI tends to pay off, read AI for small business. If you plan to bring in outside help, our guide on how to choose an AI automation agency lists the questions worth asking.
If you’d like a second opinion on your answers, or want us to run a full review across several processes with your systems and data, our readiness review service does exactly that as a fixed-scope engagement, and our broader AI consulting team can turn the results into a roadmap. Book a free initial consultation and bring your completed checklist.