AI readiness assessment

Score yourself across six dimensions, 0 to 4 points each, for a total out of 24. Every scoring level is written out below, so there is nothing to submit and no vendor tool to run — total your score, then use the band at the bottom to decide what to fix first. Data as of July 2026.

What does this measure, and how is it different from what already ranks here?

This is a self-scored, six-dimension rubric you can complete on this page — no email address, no interactive quiz, no sales call. Every framework reviewed below requires at least one of those three before it shows you a score; this one does not.

For the broader question of what AI readiness actually means and whether the right move is to build, buy, or hire, see the AI readiness overview. That page is the orientation, this page is the instrument you actually run.

Why do Microsoft, Cisco, RSM, and TDWI all use different dimensions?

Because they disagree by design. The published frameworks we could read range from four to seven dimensions, and each vendor built its assessment to lead toward its own service — a consulting engagement, a cloud platform, or a governance product.

Comparison of published AI readiness assessment frameworks
FrameworkDimensionsFormatScore visible on the page?
Microsoft745-minute interactive quizNo — must run the tool
Cisco6Interactive tool, 0–100 scoreBands are public; your score needs the tool
RSM6Four-week paid engagementNo — not self-serve
TDWI5~75-question surveyNo — gated behind a work-email form
Eide Bailly410-question snapshotPartial — short, no numeric score shown
Future Processing (5P)5Qualitative checklist articleNo numeric scoring at all

We use six dimensions below — the same count as Cisco's and RSM's — but not the same six. We fold model management (Microsoft's and RSM's seventh and sixth pillar) into infrastructure, because a company still deciding whether to hire an agent leader is rarely running enough models in production to need a separate MLOps pillar; that becomes relevant once someone already owns the function.

And we give named ownership its own dimension instead of folding it into a generic “Talent” or “Culture” bucket, which is what every framework above does. The data says ownership specifically, not skills training generally, is what predicts whether a pilot ships: the share of enterprises naming a dedicated AI agent owner went from 11% in 2024 to 56% in 2026, and that rise tracks the small share of organizations actually reaching production (Digital Applied, 2026). A pillar called “Talent” that really measures training hides that signal; we score it directly instead.

What are the six dimensions, and what does each score mean?

Score each dimension 0 to 4 using the criteria below. Be honest about where you actually are, not where the roadmap says you will be next quarter — an inflated score just produces a next action that does not match your real gap.

1. Executive sponsorship and mandate. Is there a named executive who owns AI or agent strategy, with budget and a mandate tied to a business outcome?

Scoring criteria for executive sponsorship and mandate
PointsWhat it looks like
0No executive owns AI; it is grassroots IT experimentation only.
1An executive is aware and supportive but has not allocated budget or headcount.
2Budget is approved for pilots but is not tied to a measurable business outcome.
3A written strategy exists, owned by an executive, tied to specific outcomes and a budget.
4A named leader owns AI strategy full time, reports to the CEO or board on it, and controls budget and hiring.

2. Data foundations. Is the data an agent would actually touch accurate, accessible, and governed?

Scoring criteria for data foundations
PointsWhat it looks like
0No inventory of the data agents would need; quality and access are unknown.
1Data exists but is siloed or needs manual cleaning before any use.
2Priority datasets are identified and partly cleaned; access is manual or ad hoc.
3Priority datasets are documented, access-controlled, and refreshed on a known schedule.
4Data agents touch is inventoried, quality-checked, access-governed, and its lineage is tracked.

3. Infrastructure and tooling. Can you deploy, monitor, and roll back an agent in production, not just prototype one?

Scoring criteria for infrastructure and tooling
PointsWhat it looks like
0No AI platform decision has been made; there is no sandbox to test in.
1A single-vendor pilot sandbox exists (an API key) with no path to production.
2A chosen platform supports pilots; the production deployment path is undefined.
3A production-capable stack (logging, versioning, rollback) runs at least one live agent.
4A standardized, monitored platform runs multiple agents in production against defined SLAs, including model updates.

4. Governance and risk controls. What happens when the agent is wrong, or touches sensitive data?

Scoring criteria for governance and risk controls
PointsWhat it looks like
0No AI-specific policy; agents would run under general IT policy only.
1A policy exists on paper with no enforcement or review mechanism.
2New use cases get a review gate, but live agents are not monitored afterward.
3A cross-functional process reviews new agents before launch and audits live ones periodically.
4A staffed governance function (legal, security, and business) reviews and monitors every agent against documented risk tiers.

5. Named ownership and accountability. Is one person accountable for whether this works, with the authority to fund or stop it?

Scoring criteria for named ownership and accountability
PointsWhat it looks like
0No individual owner; decisions default to whoever raises the issue that week.
1An owner is named informally but has no authority to fund or stop a project.
2An owner has authority over one function's agents, not the company-wide roadmap.
3A single leader owns the company-wide agent roadmap as a defined, if fractional, part of their role.
4A full-time leader, or a fractional one with a clear written mandate, owns agent strategy, execution, and outcomes company-wide.

6. Culture and change readiness. Will the organization actually adopt what gets built, or route around it?

Scoring criteria for culture and change readiness
PointsWhat it looks like
0No communication to staff about AI or agents; adoption would be a surprise.
1Leadership has announced intent, but no training or workflow redesign has started.
2Pilot teams are trained on specific tools; there is no company-wide change plan.
3A change plan exists with training tied to specific workflow changes.
4Staff are trained, workflows are redesigned around agent output, and adoption is tracked as a metric.

How do you total your score?

Add the six scores together for a total out of 24. There is no weighting: every dimension blocks progress equally, because a 4 on strategy does not offset a 0 on governance — it just means you are well-funded to build something you cannot yet control.

What does your score mean, and what should you do next?

A 19–24 total means you are ready to size and hire. A 12–18 total means one or two named gaps are holding you back. A 6–11 total means a single foundational dimension is likely blocking everything else. A 0–5 total means get a sponsor and an owner before spending further.

AI readiness score bands and recommended next action
ScoreBandWhat it meansDo next
19–24Deploy-readyYou have the ingredients; the constraint is sequencing, not readiness.Size the hire or the build
12–18Ready with named gapsOne or two dimensions are the actual blocker, not everything at once.Scope a fixed audit to prioritize the gap
6–11Foundational work firstData, governance, or ownership is likely blocking the rest.Fix that one dimension before adding a pilot
0–5Not yetNo one is accountable and no budget is committed.Get a sponsor and a named owner first

If you scored in the top band, size the actual cost of the hire or build with the cost estimator, which uses sourced 2026 market data for full-time, fractional, and placement pricing. If you scored in either middle band, a scoped diagnostic is usually faster than guessing which gap matters most — the Agent Readiness Audit scores your specific use cases, maps the governance gaps, and ends with a named owner and a 90-day roadmap, for a fixed price rather than an open-ended engagement. If you scored in the bottom band, the fastest fix is rarely a tool purchase: named ownership is the single trait that most separates the roughly 12% of agent pilots that reach production from the 88% that do not (Digital Applied, 2026) — see our breakdown of why AI agent pilots fail for the full picture.

Frequently asked questions

How long does this assessment take?

About 15 to 20 minutes if you score honestly across all six dimensions. It works best with input from whoever owns budget, whoever owns the data, and whoever would run point on the first agent — one person's guess is less reliable than three people's agreement.

Is a 0-24 score comparable to Cisco's 0-100 AI Readiness Index?

No. Cisco surveys thousands of companies to build percentile bands (Pacesetters, Chasers, Followers, Laggards); this is a self-scored rubric for one organization, not a benchmark against a global sample. Use this for a fast, directional read, and Cisco's or Microsoft's tool if you want a benchmarked score.

Do we need a data team to fill this out?

No, but you need someone who can honestly answer the data-foundation questions — whoever owns your data warehouse or BI stack, or whoever would be asked to build one if it doesn't exist yet. Guessing on that dimension defeats the point of the exercise.

What if we score low on every dimension?

That's common, and it's useful information rather than a bad outcome. A 0-5 total means the highest-leverage next step is getting a named executive sponsor and a single accountable owner in place — everything else on this list is premature until that exists.

Should we run this before or after hiring a Head of AI?

Before, ideally. Your two lowest-scoring dimensions usually tell you what the first hire needs to fix in their first quarter, so run this before writing the job description, not after the person starts.

Sources