Readiness for AI is something a leadership team can test on paper. An AI readiness checklist does that by asking whether the business has decided what AI should do, can reach the records it would use, has people ready to work beside it, has rules in place and has named someone to answer for it. This one sets out twelve checks. For each, you get the test it must meet, the proof another leader would open, and the mark that follows. The total describes what the firm has written down today, and it makes no forecast about a first project. The use case sets out the readiness picture for your firm more widely.
Last reviewed 2026-11-02 by Manish Sharma.
AI Readiness Checklist: Decisions, Data, People, Rules and Review
Each leadership team marks its own firm here. The exercise is a self-check, and nobody receives an audit report or a certificate at the end.
Two words need fixing first. A first AI use means one task where an artificial intelligence tool, software that reads, writes or sorts material in place of a step a person would take, does part of the work. Proof means something written that a leader can open and read.
Under Signisys’ working method, the twelve checks fall under five groups, each asking one question:
Key Takeaway The mark follows the proof, and a check stays low when nobody can open a written proof.
The grouping and each pass test come from practice and carry no standing as a standard. For comparison, the OECD AI Principles, first adopted in 2019 and revised in May 2024, list values that include transparency, security and accountability. This checklist asks what a firm has written down. A high total is no test against those principles.
AI Readiness Checklist: Telling Strong From Shaky
Under Signisys’ working method, where the proof sits and who can reach it decide the mark.
Strong (2): any member of the leadership team can open the proof today and find the pass test met. Shaky (1): someone can describe the practice, but the proof is spoken, half written, or kept where only its author can find it. Bare (0): nobody can describe the practice, and nobody can point to proof.
A spoken assurance lifts a check no higher than Shaky, whatever the seniority of the person giving it. Beside each mark, the marker notes where the proof sits. Twelve checks give a maximum of 24. There is no target score and no band, and the sheet sets no firm against another.
For example, picture an insurance broker made up for this page, with no Signisys client behind it. Its operations head marks check 1 for a plan to have AI draft renewal reminder letters. Her mark could land three ways:
- Strong: she opens a dated note naming the task and the renewals team, and the finance head has read it.
- Shaky: the leaders agreed the task in a meeting, but nobody wrote a note.
- Bare: she asks three leaders and hears three different tasks.
Which Task Would AI Take On, and to What End?
Decision and use comes first: each later check tests a task someone has named. Three checks cover it.
| Check | What another leader can open | Strong when |
|---|---|---|
| Check 1: A leader names one task for a first AI use | A short note naming the task and the team that does it today | A leader outside that team reads the note and can say what would change for the team |
| Check 2: The team head writes the result the firm wants to see | That line, in the same note | The team head can say how they would notice the result, without a number the firm has not measured |
| Check 3: A named person can stop the work | The name, and the point at which that person would stop | That person has read the line and agreed to it |
For the broker, the result in check 2 might be fewer days to answer a renewal query. That wording is enough as it names something the team head can watch for.
Picking one task first leaves room for a focused trial later. How that trial runs sits outside this page. A separate guide covers it in an order of work from a chosen task through a trial to daily use.
Where Would an AI Tool Find Its Records?
The data group asks three plain things. Do the records sit where people think, are they in fair shape, and can anyone pass them on? None of the checks says how much data is enough.
| Check | What another leader can open | Strong when |
|---|---|---|
| Check 4: The task owner lists the records the task would use and where each one sits | The list, with system or folder names | The person who looks after each place confirms the entry |
| Check 5: Someone reads a small sample of those records for gaps and mistakes | The reader’s notes | The notes name the reader and the first fix they would make |
| Check 6: The person who runs each system says whether its records can pass to another tool, and on what terms | A written answer for each system | Each answer reads yes, no, or yes on a stated condition |
For example, at the broker, check 4 might list the policy system, a shared folder of letter templates and the renewals team’s own spreadsheet.
When the task owner hands over the list, ask each system’s owner to confirm their entry then. If a name is missing, write it in before the sample is read. Note the date when each entry is confirmed.
Check 6 then puts one direct question to each system’s owner: can the tools the firm already runs work together on this task? The answer goes on file in writing, even when it is a plain no.
Who on the Staff Would Work Beside It?
People and skills is where skilling for AI readiness shows up on paper. Here the sheet names the staff who would work with the tool, and the person who teaches them.
| Check | What another leader can open | Strong when |
|---|---|---|
| Check 7: The team head names the staff who would use the tool and what each would check in its output | Names, with one line each | Those staff have read their line |
| Check 8: A manager writes a training plan: who learns what, and from whom | The plan | The plan names the people attending and the person teaching |
Keep the line in check 7 short on purpose. At the broker, a renewals clerk’s line might read: confirm the policy number and renewal date in each draft letter before it goes out.
If a line runs past one row, cut it back before staff read it. When it still needs a paragraph, compare the task with the one named in check 1.
The plan in check 8 names no course, provider or length. It names people, which is what another leader can confirm by asking them.
What Rules Would Apply From the First Day?
Rules and security covers what staff may share with an AI tool, and which laws and contracts reach the records the task would use.
| Check | What another leader can open | Strong when |
|---|---|---|
| Check 9: The firm writes which information staff may and may not put into AI tools | The written rule, and where staff find it | A member of the named team can find it without help |
| Check 10: The person responsible for legal matters lists the laws and contract terms that touch the records | The list, with the reviewer’s name | That person has signed it |
For a broker, the list in check 10 might include data protection law and the terms of client and insurer contracts. The reviewer decides what belongs on it; this page names no law in detail.
Access rules and policy deserve a deeper look, too, and a separate use case covers governing how staff use AI tools.
Who Keeps Watch Once It Is Running?
Ownership and review asks who answers for the AI use after launch, and when the firm will look at it again.
| Check | What another leader can open | Strong when |
|---|---|---|
| Check 11: The leadership team names one owner for the AI use after launch | The name, added to the note from check 1 | That owner can say whom they would tell if the output goes wrong |
| Check 12: The owner sets a review point to keep, change or stop the use, judged against the result in check 2 | The date and the result line | The team head and the person from check 3 both know the date |
Similarly, at the broker, the owner might be the renewals manager, with the operations head as the first person told about a bad letter.
Key Takeaway When a name and a date sit on the sheet, the firm knows who looks again and when.
Setting the review point before launch gives the keep-or-stop decision a date and a yardstick from the start. Without it, nothing fixes when the review takes place. What goes wrong once a trial is under way is covered in what makes a pilot lose momentum after its first demo.
AI Readiness Checklist: A Queue for the Low Marks
Write each Bare and each Shaky mark on its own line. Add the check number and the name of the leader who could produce its proof, then date the sheet.
Under Signisys’ working method, the queue runs by check number, not by group. Check 1 comes first, as checks 2 to 12 test a task someone has named. Check 11 comes next: a first use with no owner leaves nobody to act on the other lines. After those two, the remaining Bare lines follow in check order, and then the Shaky lines.
Read the total as a list of what is already written down, not as a forecast. After a change, mark the AI readiness checklist again and keep the earlier sheet beside the new one.
Laying the old sheet next to the new one shows which lines moved, and who produced the proof that moved them.
Past the Marking Sheet
Twelve checks reach only what a firm can write down about itself. They do not show whether customers or clients would accept work an AI tool has drafted, or whether the first task is worth the team’s effort. Nor do they say what a particular tool would involve, a question that comes later and sits outside the sheet. How the tool is sourced is a separate question, set out in how a firm can get an AI system made and kept running.
A talk with Signisys can help when the first lines of the queue fall in Data or in Rules and security, and nobody inside the firm can produce the proof.
Signisys provides data and AI services; its enterprise AI solution page and its data and AI service page set them out. Read the box that follows as an offer from Signisys rather than neutral advice; no outcome is promised. When you book, bring your queue sheet and the names beside its first three lines.
AI Readiness Checklist: A Leaders’ FAQ on Totals, Proof and Trials
Should we hold off on AI if our AI readiness checklist total is low?
Is the AI readiness checklist usable without a data team?
Would choosing another AI tool change the AI readiness checklist?
Which kinds of proof can lift a check to Strong?
Could a short trial lift the mark on a check?
Security Insights, Delivered.
Practical, vendor-neutral analysis of threats, tools, and architecture decisions.