An AI readiness checklist
you can tick through.
30 checks across goals, data, tools, people, risk and measuring a pilot. Each says what to look at, how to check it and who usually does it. Tick as you go, then export or print.
Free, no signup · Ticks stay in your browser
Checks done
0 / 30
- Calculated score
- 0 of 69 points
- Calculated: High checks open
- 13
- Goals and use case0/5
- Data and documents0/5
- Tools and access0/5
- People and process0/5
- Risk, privacy and checking0/5
- Measuring the pilot0/5
A checklist, not a test: nothing here looks at your business, and your ticks never leave this browser.
Brands our teams have delivered for
A free AI readiness checklist: 30 checks in 6 areas, each rated for impact and owner. Tick checks as you go. Progress is saved in your browser, and you can export a CSV or print. The tool shows a calculated score from your own ticks. Nothing is sent to us, and there is no signup.
Start with the task,
not the tool.
AI readiness is not one test with one pass mark. It is a set of questions about a single task: is it clear, is the information usable, can the tools connect, are the people ready, what happens when it goes wrong, and how will you know it helped. This list asks those questions in that order.
Every check carries an impact rating. High means a pilot is likely to stall or cause harm without it. Medium means it slows the pilot or makes the result harder to trust. Low is worth doing once the rest is done. Each check also names who usually does it: the owner, someone in operations, or someone technical.
Your ticks are saved in this browser only, so they are still here when you come back on the same device. Nothing is sent to PixelCrayons. To share progress, export the CSV and open it in Google Sheets or Excel.
AI readiness checklist (30 checks)
pixelcrayons.com/tools/ai-readiness-checklist · last reviewed 16 November 2026
30 checks,
in the order to ask them.
Work top to bottom: a pilot with no clear task cannot be rescued by better data. Tick only what is true today, and leave the rest open. Last reviewed (changelog).
Goals and use case
Do you know which task you want help with, who decides, and why? A clear task makes every later check easier.
0 of 5 done
- High impactWho usually does it: Owner
- What to look at
- Whether your goal is a single repeated task, such as sorting enquiries or drafting first replies.
- How to check
- Write the task in one sentence that a new hire would understand.
- Done looks like
- One sentence names the task, who does it now and what should change.
- High impactWho usually does it: Owner
- What to look at
- A named person who can decide, find time for the work and say stop.
- How to check
- Ask who would be called first if the pilot went wrong.
- Done looks like
- A named person has agreed to own it and knows what that means.
- Medium impactWho usually does it: Owner
- What to look at
- Which cost, delay or error you want to reduce.
- How to check
- Pick one reason, such as time, cost, speed or quality, and write it in plain words.
- Done looks like
- One reason is written down, with the way you would notice it has improved.
- Medium impactWho usually does it: Operations
- What to look at
- A task that one team can test with a few people and one source of information.
- How to check
- Check that it needs one team, one set of documents and one tool.
- Done looks like
- The pilot has one team, one source and a short list of things it will not do.
- Low impactWho usually does it: Operations
- What to look at
- Whether a template, a form change, a rule in your email tool or a setting in software you already pay for would do the job.
- How to check
- Ask a person who knows the task whether a fixed rule could handle most cases.
- Done looks like
- You know why a simpler fix would not be enough, or you have chosen to use one.
Data and documents
AI work is only as good as what it reads. These checks show whether your information is findable, current and safe to use.
0 of 5 done
- High impactWho usually does it: Operations
- What to look at
- The documents, inboxes and systems the task draws on, including the ones kept on one person's laptop.
- How to check
- List each source and the person who can open it.
- Done looks like
- A list of sources exists, each with a named person who can open it.
- High impactWho usually does it: Operations
- What to look at
- Old prices, policies or process notes that a tool would treat as true.
- How to check
- Open a small sample and compare each last-edited date with how the business works today.
- Done looks like
- Out-of-date documents are fixed, removed or marked as old.
- Medium impactWho usually does it: Technical
- What to look at
- Duplicate records, fields filled in different ways and missing values.
- How to check
- Pull a sample of records and note how often the same thing is written differently.
- Done looks like
- The common inconsistencies are listed, and the ones that matter to the task are fixed.
- High impactWho usually does it: Operations
- What to look at
- Customer details, staff records, contracts, payment or health information inside the sources.
- How to check
- Scan each source for these kinds of data and note where they sit.
- Done looks like
- Each source is marked as holding sensitive data or not, and the pilot leaves the sensitive ones out.
- Medium impactWho usually does it: Owner
- What to look at
- Licensed material, client files and third-party terms that limit reuse.
- How to check
- Read the terms that cover each source, and ask your legal adviser if any are unclear.
- Done looks like
- Each source is cleared for the pilot, or left out.
Tools and access
Can the tools you use today be connected safely, and can the pilot run without risk to live records?
0 of 5 done
- Medium impactWho usually does it: Operations
- What to look at
- Email, CRM, help desk, spreadsheets, calendars and anything else the task touches.
- How to check
- Walk through the task once and write down each system you open.
- Done looks like
- A list names each system, who administers it and what the task needs from it.
- High impactWho usually does it: Technical
- What to look at
- Whether each tool allows a separate account, read-only access or limited permissions for a pilot.
- How to check
- Check each tool's user and permission settings, or ask its administrator.
- Done looks like
- The pilot has its own limited access, and no one shares a personal login.
- Medium impactWho usually does it: Technical
- What to look at
- Whether each tool has an official connection or open interface, or needs manual exports.
- How to check
- Read each vendor's documentation or ask the vendor. Treat anything you cannot confirm as not connected.
- Done looks like
- Each tool is marked as connected, manual or unknown, with a source for the answer.
- Medium impactWho usually does it: Technical
- What to look at
- A test copy, a sandbox or a set of sample records, so the pilot never touches live ones.
- How to check
- Ask whether you can run the pilot on copies, and who would set that up.
- Done looks like
- There is a test copy or sample set, and live records are out of reach of the pilot.
- Low impactWho usually does it: Operations
- What to look at
- AI features inside software you already pay for, and any AI tools staff use on their own.
- How to check
- Check each product's settings and ask staff, without blame, which tools they already use.
- Done looks like
- You have a list of AI tools in use and a written rule on what staff may paste into them.
People and process
A pilot changes how people work. These checks make sure the people and the steps are ready before the tool arrives.
0 of 5 done
- High impactWho usually does it: Operations
- What to look at
- The staff who do it today, and what they find hard or slow.
- How to check
- Ask two or three of them to walk you through a normal day of the task.
- Done looks like
- The people who do the task have shaped the pilot and have said what worries them.
- High impactWho usually does it: Operations
- What to look at
- Each step, who does it, and the rules for odd cases.
- How to check
- Write the steps as a numbered list, then ask the person who does the task to correct it.
- Done looks like
- A numbered list of steps, checked by the person who does them.
- Medium impactWho usually does it: Operations
- What to look at
- Hours set aside for testing and for giving feedback during the pilot.
- How to check
- Ask the team lead to block the time in the diary before the pilot starts.
- Done looks like
- Test time is in diaries, and the usual work is covered while it happens.
- Medium impactWho usually does it: Owner
- What to look at
- Worries about jobs, about being watched and about extra work.
- How to check
- Tell the team in plain words what the pilot will and will not do, and invite questions.
- Done looks like
- The team has heard the plan from the owner and has had a chance to ask questions.
- Medium impactWho usually does it: Owner
- What to look at
- Who watches, updates and fixes the system once it is live.
- How to check
- Decide whether your own team, a managed plan or a partner will do it.
- Done looks like
- A named person or team is responsible after the pilot, and the cost of that is known.
Risk, privacy and checking
AI gets things wrong sometimes. These checks decide where a person must look, and what happens when a mistake gets through.
0 of 5 done
- High impactWho usually does it: Owner
- What to look at
- The laws, contracts and industry rules that cover your data. They depend on your country, your sector and your customers.
- How to check
- Ask whoever handles privacy or legal advice for you, and write down the rules they name.
- Done looks like
- A short list of the rules that apply, confirmed by someone qualified to say so.
- High impactWho usually does it: Operations
- What to look at
- Which results go to customers or change records without anyone looking first.
- How to check
- Mark each output as sent automatically or checked first. Check first by default.
- Done looks like
- Every output that reaches a customer or changes a record has a named checker.
- High impactWho usually does it: Owner
- What to look at
- The worst likely mistake, and who it would affect.
- How to check
- Write down three ways it could go wrong and how each would be caught and fixed.
- Done looks like
- Three failure cases are written down, each with a way to catch it and a way to fix it.
- Medium impactWho usually does it: Operations
- What to look at
- A clear route for a customer or colleague to reach a human when the tool cannot help.
- How to check
- Try the route yourself, as a customer would, and note how long it takes.
- Done looks like
- You tested the route to a person and it works without a dead end.
- Low impactWho usually does it: Technical
- What to look at
- Whether you can see what the tool did, what it was given and who approved changes to it.
- How to check
- Check whether the tool keeps a history. If not, decide where a person will keep notes.
- Done looks like
- A record exists of what the tool did and who changed its settings.
Measuring the pilot
Without a starting point you cannot tell whether the pilot helped. Set these before the first test.
0 of 5 done
- High impactWho usually does it: Operations
- What to look at
- The time taken, the number of errors or the wait for a reply, before anything changes.
- How to check
- Time a handful of real cases, or count errors across a normal stretch of work.
- Done looks like
- A starting measure is written down, with the date and the way you took it.
- High impactWho usually does it: Owner
- What to look at
- The change that would justify carrying on.
- How to check
- Write the target before the pilot starts, and agree it with the owner.
- Done looks like
- A target is written and agreed before the first test.
- Medium impactWho usually does it: Operations
- What to look at
- Real past examples where the right answer is already known.
- How to check
- Pick varied examples, including awkward ones, and have a person write the correct result for each.
- Done looks like
- A set of past cases with correct answers, kept apart from anything the tool has seen.
- Medium impactWho usually does it: Owner
- What to look at
- A date when someone decides to stop, change or widen the pilot.
- How to check
- Put the date in the diary and invite the owner and the team lead.
- Done looks like
- A review is booked, and the people who decide are invited.
- Low impactWho usually does it: Owner
- What to look at
- Software, set-up, staff time for checking and the time spent fixing mistakes.
- How to check
- List each cost, including the hours your team will spend, and compare it with the target.
- Done looks like
- A cost list that includes staff time, beside the target it must meet.
How the score is worked out
The tool adds up points for the checks you tick. A High check is worth three points, a Medium check two and a Low check one. It also counts the High checks still open.
The score is arithmetic on your own ticks, nothing more. It is not a test result and nothing here compares you with other businesses.
There is no standard pass mark. Our rule of thumb is to wait until every High check in goals, data and risk is ticked. A High check still open in one of those areas is the first thing to fix, whatever the total says.
What a filled-in list
looks like.
Example: an invented online shop with a three-person support team, wanting help with order questions. These are made-up answers to show how the list reads. They are not a client and not a result.
- Example · Goals and use case
One task is named, not just "use AI"
The task is to sort order questions and draft first replies. The support lead owns it.
Done - Example · Data and documents
The documents are up to date
The returns policy document was last edited long before the shop changed its courier. It needs a rewrite before the tool reads it.
Open - Example · Data and documents
Personal and confidential data is marked
Order records hold names and addresses. The pilot reads policy documents and past replies only, with those details removed.
Done - Example · Tools and access
Access can be limited
The help desk allows a separate account with limited permissions. The pilot uses it.
Done - Example · People and process
The current steps are written down
The three people who answer questions wrote the steps together and corrected each other's lists.
Done - Example · Risk, privacy and checking
The privacy and legal rules that apply are known
No one has yet asked the shop's adviser which privacy rules apply to customers in other countries.
Open - Example · Risk, privacy and checking
A person checks the output that matters
Every drafted reply is read and sent by a person. Nothing is sent automatically.
Done - Example · Risk, privacy and checking
You know what happens when it is wrong
The three failure cases are not written down yet. The team has only talked about them.
Open - Example · Measuring the pilot
You have measured how the task works today
The team timed a sample of real replies and counted the questions that needed a second reply.
Done - Example · Measuring the pilot
You know what better looks like
No target is set yet. The team agrees that it must be written before the first test.
Open
16 of 30 checks done, 41 of 69 points.
High checks still open: The documents are up to date; The privacy and legal rules that apply are known; You know what happens when it is wrong; You know what better looks like.
Example data, invented for this page. The score line is calculated from the ticked checks using the weights above. In this example the shop would fix the open High checks before starting the pilot.
Last reviewed
- 16 November 2026
First edition: 30 checks in 6 areas, with a calculated score and a CSV export.
- AI readiness assessment The paid version: we look at your tasks, data and tools, and send a written plan.
- Sample AI readiness report An invented example of the report the assessment ends with. Every section is marked Example.
- AI ROI calculator Type your own numbers to see the hours freed and the share needed to cover a plan.
- Free AI automation review Name up to three repeated tasks and get a ranked shortlist of what to automate first.
- AI consulting Advice on where AI fits in your business and how to start safely.
- AI services Everything we build and run in AI, grouped by the job it does.
- AI marketing Using AI inside your marketing, with a person checking what goes out.
Frequently
asked.
It is a list of questions that show whether your business can start an AI project with a fair chance of it working. The questions cover the task you want help with, your data, your tools, your people, your privacy duties and how you will measure the result. The answers show what to fix before you spend money on a pilot.
Most people can read it through in one sitting. Answering every check honestly takes longer, because some answers need a colleague, a document or a look at a system. Tick what you know now, export the list, and come back to the rest. Your ticks stay in this browser until you clear them.
There is no standard pass mark, and any tool that claims one is guessing. Look at the open High checks instead. If every High check in goals, data and risk is ticked, you are in a fair position to run a small pilot. If one is open, fix it first, whatever the total says. The score here is calculated from your ticks and nothing else.
Yes. There is no form, no email and no signup. You can tick the checks, export a CSV and print the page without giving us any details. The paid readiness assessment is a separate service, and you do not need it to use the list.
No. Your ticks are saved in your own browser and are never sent to us. If you clear your browser data or use another device, they will not be there. Export the CSV if you want to keep a copy or share it with your team.
Want help with
the open checks?
The AI readiness assessment looks at your tasks, data and tools and sends a written plan. Or start with the free review: describe up to three repeated tasks and a person sends a ranked shortlist.
Free review · No retainer required · Yours to keep
Last updated















