Good: repetitive, high-volume work
Tasks that recur many times a week give a pilot plenty of examples and a clear before and after.
AI readiness guide
AI Readiness Assessment: A First-Pass Checklist for Founders
An AI readiness assessment is a structured look at whether a company can use AI well on real work. It can cover areas such as processes, data, people, tools, governance and priorities. Below, I give a checklist you can run yourself and a way to pick first pilots.
On this page
- What an AI readiness assessment covers
- A first-pass AI readiness checklist
- How to read your checklist results
- How to pick your first AI pilots
- Good and poor candidates for a first pilot
- Do you need clean data before starting an AI pilot?
- What an AI readiness assessment cannot do
- Where AI opportunity assessment appears in my work
- Frequently asked questions
What an AI readiness assessment covers
This article treats an AI readiness assessment as a review of six areas: processes, data, people, tools, governance and priorities.
The purpose is to find where AI can help now, where preparation is needed first and where AI should wait. The output can be a short list of candidate pilots and a short list of gaps.
An AI opportunity assessment is a close relative. Readiness asks whether the company can use AI well. Opportunity assessment asks where AI should be used first.
AI readiness for a small business can start with a simple review. A founder-led company can run a useful first pass itself. The main requirement is honest answers from the people who do the work.
Involve those people directly. Include team leads, the owner of each process, whoever manages data and systems, and a leader who can decide. Each brings a different view of what is ready.
That first pass suits the pattern I describe in AI without a strategy. Experiments run everywhere, and nobody knows where to use AI.
A first-pass AI readiness checklist
Answer each question for your company. Ask the person closest to each process, and note the evidence behind each rating. Rate each area ready, partly ready or not yet, based on what you can show rather than what you hope.
| Area | Questions to answer | Warning sign |
|---|---|---|
| Processes | Which tasks are documented, repeated often and owned by someone? Where does work wait or get re-entered? | Processes live inside people's heads. |
| Data | Where does the information for each task live? Who can access it, and how current is it? | Key information sits in personal spreadsheets or inboxes. |
| People | Who would use an AI tool, and who would check its output? Do they have time and interest? | Nobody owns the pilot, or staff fear being replaced. |
| Tools | Which systems hold the work today? Can they connect to anything new? | Existing systems are used unevenly, so the data may be incomplete. |
| Governance | Is there a named owner, an approved tool list and a set of data rules? | Employees use random tools with no rules. |
| Priorities | Which business problems matter most this quarter? Which of them involve repeated work? | AI interest has no link to business goals. |
How to read your checklist results
Ready areas can support a pilot now. Partly ready areas can support a pilot if someone owns the gap. Areas marked not yet need work before a pilot touches them.
Governance deserves early attention, because gaps there can create risk before any pilot starts. The AI governance guide gives a starter structure.
Process gaps deserve the same attention. In my approach, a problem that looks like a technology problem may actually be a process problem. Applying AI to a poorly designed process can speed up the wrong work.
Give each gap an owner and a next step. A gap without an owner tends to stay a gap.
How to pick your first AI pilots
A first pilot should be small, useful and easy to judge. These steps show where to start with AI by narrowing a long list of ideas to one or two.
-
List candidate tasks
Ask each team where its time goes. Note tasks that are repetitive, high in volume and based on clear rules. Capture them in one list.
-
Filter for repeatable, high-volume work
Frequent work gives a pilot many examples to test against. One-off tasks give too few.
-
Check that you can measure it
Record how the task runs today, before any change. Note time spent, error rate and turnaround. Without a baseline, you can't tell whether the pilot helped.
-
Check data and risk
Confirm the data is available and permitted under your data rules and contracts. Prefer tasks where a person reviews the output before it reaches a customer.
-
Score the shortlist
Rate each candidate on volume, clarity of rules, measurability, data readiness and risk. Compare candidates side by side. Prefer the one that scores steadily across the board over one that excels on a single measure.
-
Assign an owner and a small scope
Give the pilot one owner and a narrow boundary. A single team, one process and a named reviewer make a workable size.
-
Define success and a stop rule
Write down what result counts as success and when you will stop or adjust. Review the pilot at a set date, then decide to scale, change or end it.
Good and poor candidates for a first pilot
Use these patterns to sort your candidate list.
Good: rules-based work with clear inputs
Work with a clear input and a clear expected output is easier to test and to review.
Good: output a person can check
A reviewer can catch errors before they reach a customer or a decision.
Good: a contained scope
A single team and a single process keep a pilot easy to run and easy to stop.
Poor: high-stakes decisions
Decisions that affect a person's job, safety, legal position or large sums need more care than a first pilot offers.
Poor: undocumented, unowned work
If nobody can describe how a task runs today, redesign the process first. Then consider AI.
Poor: tasks tied to messy or restricted data
Data that is scattered, out of date or off limits under your rules will stall a pilot.
Poor: a company-wide rollout as the first step
A first pilot that spans every team is hard to run, hard to measure and hard to stop.
Do you need clean data before starting an AI pilot?
You need clean data only for the task an AI pilot covers. That data must be findable, accessible and good enough to test against. That is a lower bar than cleaning every dataset in the company.
Pick a pilot whose data you can inspect. Sample real records, look for gaps and duplicates, and note what would need fixing. If the task depends on records that are missing or unreliable, choose a different pilot or fix the source first.
Data problems are often a sign that process work should come first. Where the same fact is typed into several places, the records can drift apart.
What an AI readiness assessment cannot do
An AI readiness assessment has limits, and they are worth stating plainly. The list below covers the main ones.
- It can suggest where value may lie but cannot confirm a return before a pilot runs.
- It is no substitute for testing an idea on real work.
- It cannot fix a broken process or an unclear owner.
- It is only as accurate as the answers, so the people who do the work should give them.
- It goes out of date as tools, data and priorities change.
- It cannot show that AI is the right answer to every problem, because some tasks suit simpler fixes.
- Requirements come first, so the assessment cannot pick your tools for you.
Where AI opportunity assessment appears in my work
My Fractional COO + AI Transformation engagement includes an AI opportunity assessment, alongside AI implementation, workflow automation and AI governance. The AI transformation service sets out the full list.
Separately, in my 90-Day Scale-Up Operating System, I assess AI opportunities in Phase 1 (Diagnose, weeks 1–2). Phase 2 (Design, weeks 3–4) creates an AI opportunity map. Phase 3 (Implement, weeks 5–12) covers AI pilots and automation.
The checklist above is a first pass you can run yourself. If you want to talk through where AI fits in your company, start a conversation.
Frequently asked questions
What is an AI readiness assessment?
An AI readiness assessment is a structured review of whether a company can use AI well. It often covers processes, data, people, tools, rules and priorities. Its purpose is to show where AI can help and what needs fixing first. You can run a first pass yourself with the checklist above.
How do you pick the first AI project?
Pick the first AI project from work that is repetitive, high in volume and measurable. A person should be able to review the output. Give the project one owner, a narrow scope, a baseline and a stop rule.
What is an AI opportunity map?
In general terms, an AI opportunity map is a view of where AI could support the work. My 90-Day Scale-Up Operating System lists an AI opportunity map among its Phase 2 outputs (Design, weeks 3–4). Phase 1 assesses AI opportunities first.
Do we need clean data before starting?
Clean data is required only for the task a pilot covers. That data must be findable, accessible and reliable enough to test against. Inspect a sample of real records before you commit.
How is an AI readiness assessment different from an AI opportunity assessment?
An AI readiness assessment asks whether a company can use AI well. An AI opportunity assessment asks where AI should be used first. The two can overlap. My Fractional COO + AI Transformation engagement includes an AI opportunity assessment.
How often should you repeat an AI readiness assessment?
Repeat an AI readiness assessment when tools, data or priorities change, and after each pilot. A quick review of the checklist can be enough. Update the gap list and the pilot list each time.
Keep reading
Talk about where AI fits
If your company is interested in AI but has not operationalized it, let's talk about where to begin.
Book a conversation