AI Readiness Assessment: A 15-Minute Scorecard and Maturity Model for Executives
Guide · · Published October 2026 · 10 min read
By Ashwin Gopinath, CEO
TL;DR
- An AI readiness assessment scores how prepared your company is to put AI into real work. This one covers seven dimensions, each scored 1 to 4, and takes about 15 minutes for one leader to complete.
- Add the seven scores for a total between 7 and 28. The total places you on a five-stage AI maturity model and tells you what to fund next.
- Data and context access is the dimension to watch. When it scores low, AI tools cannot see the email, chat, meeting and ERP records that explain how the business actually runs.
- Sentra raises that score. It connects those sources into one permission-aware company memory that ChatGPT, Claude, Copilot, Gemini and internal agents can all use, and a named person approves anything Sentra raises before it reaches anyone else.
See what your leadership team would know before Monday's review.
Book a demoWhat an AI readiness assessment is
An AI readiness assessment is a structured self-review of whether your company has the ownership, data, controls, skills, use cases, budget and measurement needed to get value from AI. It produces a score for each dimension and an overall score, so you can see which gap is blocking you rather than arguing about AI in general.
AI readiness is different from AI interest. A company can have hundreds of people using chat assistants and still score low, because nobody owns the results, the assistants cannot see company data, and no one measures what changed. The assessment below is built to surface exactly that kind of gap.
The cost of skipping it is documented. Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data, and found that 63% of organizations either do not have or are unsure if they have the right data management practices for AI (gartner.com). Data readiness for AI is one of the seven dimensions here, and it is the one to score most strictly.
How to run the assessment in 15 minutes
1. Pick one accountable leader to score, usually the COO, CIO or Head of AI. Scoring by committee drifts toward the middle.
2. Read each dimension and choose the level that matches what is true today, not what is planned. If you sit between two levels, take the lower one.
3. Write one line of evidence next to every score: a document, a dashboard, a named owner. A score with no evidence counts as 1.
4. Add the seven scores and find your band in the score table further down.
5. Share the scored table with the executive team and ask each person to challenge one score. Disagreement is useful data.
The AI readiness checklist: seven dimensions, four levels each
1. Strategy and ownership
| Level | What it looks like |
|---|---|
| 1 | AI is discussed in leadership meetings but has no owner, budget line or written goals. |
| 2 | A named owner exists, but goals are activity based, such as "roll out assistants to all staff". |
| 3 | An executive owns AI with a budget, and each initiative has a business owner and a target metric. |
| 4 | AI goals sit in the operating plan. Owners report results in the same review as revenue and cost. |
2. Data and context access
| Level | What it looks like |
|---|---|
| 1 | AI tools see only what a user pastes in. Company data stays in its source systems. |
| 2 | One or two connectors exist, usually documents or a wiki. Email, chat and meetings are excluded. |
| 3 | Most core systems are connected, including CRM and ERP, with access that follows source permissions. |
| 4 | AI can see systems of record and conversations together, knows which facts are current, and every answer cites its source. |
3. Security and governance
| Level | What it looks like |
|---|---|
| 1 | No AI policy. Staff use personal accounts for work tasks, which is the pattern described in our guide to shadow AI. |
| 2 | A written acceptable use policy exists, but tools are not approved centrally and nothing is logged. |
| 3 | Approved tools only, with SSO, data processing agreements and a review step for new use cases. |
| 4 | Every AI answer and action respects existing permissions, high-impact actions need a named approver, and activity is auditable. |
Two references help at levels 3 and 4. The NIST AI Risk Management Framework is voluntary guidance released in January 2023, organized around four functions: Govern, Map, Measure and Manage. ISO/IEC 42001:2023 specifies requirements for establishing, implementing, maintaining and continually improving an AI management system, for companies that want a certifiable standard.
4. People and skills
| Level | What it looks like |
|---|---|
| 1 | No training. A few enthusiasts teach themselves. |
| 2 | Optional training on prompting. No role-specific guidance. |
| 3 | Role-based training for every function using AI, with named champions in each team. |
| 4 | Managers redesign work around AI, review its output as part of the job, and hiring plans account for it. |
If you operate in the EU, this dimension is not optional. Article 4 of the EU AI Act requires providers and deployers of AI systems to take measures to support the AI literacy of their staff, and it has applied since 2 February 2025 (artificialintelligenceact.eu).
5. Use-case pipeline
| Level | What it looks like |
|---|---|
| 1 | Ideas arrive by email and are judged by enthusiasm. |
| 2 | A list of use cases exists, but none have a baseline metric or owner. |
| 3 | Use cases are scored on value, feasibility and risk, and two or three are in production with owners. |
| 4 | A steady pipeline moves from idea to pilot to production on a set cadence, and failed pilots are stopped quickly. |
6. Tooling and cost
| Level | What it looks like |
|---|---|
| 1 | Tools are bought team by team. Nobody knows total AI spend. |
| 2 | Seat licences are tracked, but usage-based costs such as API tokens are not. |
| 3 | One owner tracks seats, tokens and integration spend, with a monthly cost report. |
| 4 | Cost per outcome is known for each use case, and the context sent to models is managed to keep token spend down. |
For the token side, our agent token cost calculator gives a fast estimate, and the best tools to reduce LLM token costs covers the options.
7. Measurement
| Level | What it looks like |
|---|---|
| 1 | Success is judged by anecdotes and adoption counts. |
| 2 | Usage is measured, but no baseline was taken before rollout. |
| 3 | Each use case has a before and after metric, such as hours, error rate or cycle time. |
| 4 | Results are measured against a control group and reported in financial terms. See our guide to AI ROI. |
AI readiness scoring table
Copy this table into a document and fill it in. The example rows show a typical mid-sized company partway through its first year.
| Dimension | Score (1 to 4) | Evidence | Owner | Next gap to close |
|---|---|---|---|---|
| Strategy and ownership | 3 | AI line in FY plan | COO | Add metrics to ops review |
| Data and context access | 1 | Assistants see pasted text only | CIO | Connect CRM, ERP, email and chat |
| Security and governance | 2 | Policy doc, no logging | CISO | Approved tool list, SSO, approvals |
| People and skills | 2 | One prompting workshop | Head of People | Role-based training |
| Use-case pipeline | 2 | Spreadsheet of 30 ideas | Head of AI | Score and pick three |
| Tooling and cost | 2 | Seat counts only | Finance | Track token spend |
| Measurement | 1 | Adoption dashboard | Head of AI | Baselines before next pilot |
| Total | 13 |
What your total score means and what to do next
| Total | Stage | What to do in the next 90 days |
|---|---|---|
| 7 to 10 | 1. Exploring | Name one executive owner, write an acceptable use policy, approve one assistant for everyone, and pick a single use case with a clear owner. |
| 11 to 14 | 2. Experimenting | Take baselines, connect the two systems your top use case depends on, and stop any pilot without an owner. |
| 15 to 19 | 3. Operationalizing | Move two use cases into production, add role-based training, and start tracking seats and tokens in one report. |
| 20 to 24 | 4. Scaling | Give AI access to systems of record and conversations under source permissions, add approval steps for agent actions, and report results in financial terms. |
| 25 to 28 | 5. Embedded | Measure cost per outcome, retire tools that duplicate each other, and let agents own recurring work with named human approvers. |
The example company above scored 13, so it sits in Stage 2. Its fastest gain is data and context access, because every later use case depends on it.
The AI maturity model: five stages
The total score maps to an AI maturity model. Use it to explain the plan to a board in one slide.
| Stage | How AI is used | Typical blocker | Sign you have moved up |
|---|---|---|---|
| 1. Exploring | Individuals try chat assistants on their own | No owner or policy | One approved tool and one owner |
| 2. Experimenting | Pilots in a few teams | Pilots cannot see company data | A pilot with a baseline and a result |
| 3. Operationalizing | Two or three use cases in production | Context is fragmented across systems | Use cases share the same connected data |
| 4. Scaling | AI across functions, first agents taking actions | Permissions and approvals for agents | Agents act only with named approval |
| 5. Embedded | AI owns recurring work, people review and decide | Cost and duplication | Cost per outcome falls each quarter |
Plan for at least two quarters per stage. The risk in jumping from Stage 2 to Stage 4 is agents that act without the context or permissions to act correctly.
Why data readiness for AI decides how far you get
An AI readiness framework that only asks "is our data clean" misses the larger problem. The facts that explain how a company runs are spread across systems that were never designed to be read together. The ERP holds the invoice. The reason for the discount sits in a sales call. The decision to change a supplier sits in a chat thread.
When an assistant sees only documents, it answers from whatever document is closest to the question, even if that document is out of date. That is why companies at level 1 or 2 on this dimension see confident answers that are wrong. Our explainer on why RAG fails covers the mechanism.
Three checks tell you where you stand on data readiness for AI:
- Coverage. Can your AI see email, chat, meetings and documents alongside CRM, ERP and HRIS? Or only one of those groups?
- Permissions. Does every answer respect what the asking person can see in the source system? If not, security will block wider rollout, and should.
- Currency. Can the system tell a fact that is true now from one that was true last quarter? Without this, agents restate superseded plans as current.
Where Sentra fits
Sentra raises the data and context access score without replacing the AI tools you have already approved. It connects email, chat, meetings, documents, ERP, CRM, HRIS and BI through 200+ integrations, and works out what is true now, why it changed, and who is working on it. It records when a fact became true and when it stopped being true, so an agent does not present last quarter's plan as this quarter's.
That memory is available through REST and MCP, so ChatGPT, Claude, Copilot, Gemini and internal agents all draw on the same company context instead of each building its own partial copy. Answers and actions respect each source's existing permissions, checked at query time, and a failed check counts as no access.
Sentra also acts first. It finds what is going wrong across systems, such as a stalled dependency or a figure that contradicts what was agreed in a meeting, works out who should act, and tells them. A named person approves anything Sentra raises before it reaches anyone else, which covers part of the security and governance dimension too. Sentra holds SOC 2 Type II and ISO 27001, and deploys in the cloud, in an isolated VPC, or air-gapped on premises.
At Lenskart, time to root cause fell from 3 to 4 days to about 30 minutes.
If your assessment shows data and context access as the lowest score, book a demo at /demo.
How to start after you score
1. Fix the lowest score first. A Stage 2 company with a 1 on data access will not reach Stage 4 by buying more seats.
2. Pick one use case that crosses systems. Month-end variance analysis or a weekly executive briefing are good tests, because the numbers and the reasons sit in different places.
3. Take the baseline before you change anything. Hours, errors, cycle time. Without it, the next assessment cannot show progress.
4. Re-score every quarter. Keep the evidence column. A score that rises without new evidence did not rise.
For a wider view of how agents fit into company operations, see AI agents for business.
FAQ
What is an AI readiness assessment?
An AI readiness assessment is a self-review that scores a company on the conditions AI needs to deliver value: ownership, data access, governance, skills, use cases, cost control and measurement. It shows which gap to close first.
How long does an AI readiness assessment take?
The version on this page takes about 15 minutes for one leader to score, plus a 30 minute executive review to challenge the scores. A deeper version adds interviews with each function owner.
What is a good AI readiness score?
On this seven-dimension scale, 20 or above means you are ready to scale AI across functions. Below 15, focus on ownership, data access and baselines before adding tools or agents.
What is the difference between an AI readiness assessment and an AI maturity model?
The assessment is the scoring exercise. The AI maturity model is the five-stage ladder the score places you on, which is useful for explaining the plan to a board.
What does data readiness for AI mean?
It means AI systems can reach the data they need, with permissions enforced, and can tell current facts from outdated ones. Clean tables alone are not enough if the reasons behind the numbers sit in email, chat and meetings.
Which frameworks should an AI readiness framework align with?
Many companies map governance to the NIST AI RMF functions of Govern, Map, Measure and Manage, and use ISO/IEC 42001 when they want a certifiable AI management system. EU deployers also need to meet the AI literacy duty in Article 4 of the AI Act.
Who should own AI readiness in a company?
One executive, usually the COO, CIO or a Head of AI, with a business owner for each use case. Shared ownership across a committee tends to leave no one accountable for results.