Best AI Readiness Assessment

AI readiness checklist for mid-sized companies: 24 checks in six areas

The published readiness frameworks agree on more than their pillar names suggest. Reduced to what a company of a few hundred people can check itself in two weeks, they come to 24 questions.

Contents
  1. The problem with most checklists
  2. What the frameworks have in common
  3. The checklist
  4. How to score it
  5. A two-week plan to run it
  6. Questions buyers ask
  7. Sources

The problem with most checklists

A mid-sized company usually has no AI team and no time for a framework with dozens of indicators. Its staff are often ahead of it: in Microsoft’s 2024 Work Trend Index, 78% of people using AI at work brought their own tools⁠1. The leadership team needs a short list it can answer from what it already knows, that shows where the gaps are before anyone signs a contract.

The published frameworks are a good source for that list, but each is written by an organisation with something to sell. Cisco gives infrastructure a pillar of its own⁠2. Microsoft adds model management⁠3. Consultancies organise theirs around what they deliver.

What the frameworks have in common

FrameworkAreas
Cisco AI Readiness IndexStrategy, infrastructure, data, governance, talent, culture; 49 indicators⁠24
Microsoft AI Readiness AssessmentBusiness strategy, AI governance and security, data foundations, AI strategy and experience, organisation and culture, infrastructure for AI, model management⁠3
RSM AI readiness assessmentStrategy and vision, data foundation, organisation and culture, technical infrastructure, governance and compliance, model management⁠5
Beezwax AI readiness assessmentData infrastructure, process maturity, technical readiness, team capability, strategic alignment⁠6

Strategy, data, technology, governance and people appear in every framework that names them. Process appears explicitly only in Beezwax’s, yet it decides whether the others pay off: MIT’s research found the largest returns in back-office automation, and traced most failures to tools that do not adapt to how work is done⁠7. The checklist below keeps all six.

The checklist

Answer each item yes or no. A no is not a failure; it is the next piece of work.

Strategy

  1. There is one written business goal for AI, tied to a number such as hours, cycle time, cost or revenue.
  2. A named executive owns that goal.
  3. The ten most repetitive workflows are listed, with hours per week for each.
  4. Two or three of them are chosen as first candidates, with the reason written down.

Data

  1. For each candidate workflow, we know which systems and files hold the information it needs.
  2. Customer, product and supplier records are not duplicated across systems, or the duplicates are known.
  3. The information the workflow depends on is in a system, not in inboxes and personal spreadsheets.
  4. Someone is responsible for the quality of each core data set.

Systems

  1. The CRM, ERP or accounting system, email and file storage each offer an API or a supported connector.
  2. We know which systems cannot be connected, and what that rules out.
  3. Logins for business software are managed centrally, so access can be granted and withdrawn.
  4. There is a list of the AI tools and subscriptions in use, approved or not.

Governance

  1. A one-page AI policy names the approved tools.
  2. The policy says which data never goes into an AI tool: customer data, contracts under NDA, pricing, personnel files.
  3. It names who approves exceptions.
  4. For each role and system, it is decided what AI may read, what it may draft and what needs a named approver.

People

  1. Most of the leadership team uses an approved AI tool in its own work every week.
  2. Staff in the candidate workflows know about the plan and have been asked how the work is really done.
  3. There is a plan to train the people whose work will change.
  4. Someone inside the company will own each workflow once it runs.

Process

  1. Each candidate workflow has a baseline: hours, error rate or cycle time, measured before anything changes.
  2. Each workflow has been reviewed for steps that should be removed before any are automated.
  3. We have agreed how long a first workflow runs before it is judged.
  4. We have agreed what result would justify extending it.

How to score it

Count the yes answers in each area rather than in total. Cisco’s index shows why: the best-prepared companies are not slightly better everywhere, they are far ahead in particular areas, such as fully centralised data (76% against 19% overall) and comprehensive change-management plans (91% against 35%)⁠8. One empty area can stop a project that every other area supports.

As a rule of thumb, an area with fewer than two yes answers is a foundation gap to close before buying tools. If governance is one of them, close it first: once AI can act in company systems, the permission map in item 16 is what keeps it inside the lines. The eight-question readiness check on the home page gives the same reading in three minutes, with a next step for each gap.

A two-week plan to run it

  1. Day 1. The executive who owns the AI goal circulates the checklist to the heads of finance, operations and IT, and names the candidate workflows.
  2. Days 2 to 7. Each head answers the areas they own and attaches evidence: the workflow list, the system inventory, the policy draft.
  3. Days 8 to 10. Measure baselines for the two or three candidate workflows.
  4. Day 10. The leadership team reviews the answers together and agrees on the weakest area.
  5. Days 11 to 14. Close what can be closed in a week (the policy, the owner, the tool list) and decide whether the remaining gaps need an outside assessment. The ranking compares the options, from free tools to fixed-fee reviews.

At the end of two weeks the company has an honest picture, a short list of first workflows with baselines, and a policy its staff can follow. If it then hires an assessor, it pays for analysis rather than for collecting facts it already had.

Questions buyers ask

What is on an AI readiness checklist?

The items fall into six areas. Strategy, a written goal and a list of workflows. Data, where the information lives and whether it is clean. Systems, whether core software can be connected. Governance, approved tools, data rules and approval rights. People, whether leaders use AI and staff are trained. Process, baselines and a way to measure results.

Who should fill in an AI readiness checklist?

The executive who will own the AI goal, with the heads of finance, operations and IT. Each area has a natural owner, and the answers are more accurate when the person who runs the work answers for it.

Can a company assess its own AI readiness?

Largely, yes. Most items on the checklist can be answered internally in two weeks. An outside assessment adds interviews, a review of the actual data and a prioritised plan, and is worth its fee once the company is about to commit budget.

Sources

  1. AI at Work Is Here. Now Comes the Hard Part. Microsoft WorkLab (Work Trend Index), 8 May 2024. Accessed 19 Nov 2026.
  2. AI Readiness Index Assessment Tool. Cisco. Accessed 19 Nov 2026.
  3. AI Readiness Assessment. Microsoft Learn. Accessed 19 Nov 2026.
  4. AI Readiness Index: methodology. Cisco. Accessed 19 Nov 2026.
  5. AI readiness assessment. RSM US. Accessed 19 Nov 2026.
  6. AI Readiness Assessment. Beezwax. Accessed 19 Nov 2026.
  7. MIT report: 95% of generative AI pilots at companies are failing. Fortune, 18 Aug 2025. Accessed 19 Nov 2026.
  8. Cisco AI Readiness Index. Cisco. Accessed 19 Nov 2026.