Rate 20 statements across strategy, data, infrastructure, security and governance, people, and operations. You get a score out of 100 for each area, your place on a four-level maturity scale from Exploring to Operating, and the gap to close first, with the templates and tools for each one.
Answer for the organisation or business unit you are assessing, as it is today rather than as planned. If several people answer, compare the area scores: disagreement is itself a finding.
Your overall AI readiness
Free DOCX template pack
Six editable templates for the gaps above: governance policy, incident response plan, risk register, model card, use case intake form and this readiness checklist.
The areas follow the order in which gaps usually stop AI projects: strategy decides which work is worth doing, data and infrastructure decide whether it can be built, governance decides whether it may go live, and people and operations decide whether it lasts.
Whether AI work is tied to business outcomes, ranked, and measured from a baseline, rather than started wherever enthusiasm happens to be.
If it scores low: Write each candidate on one page with its owner, data, risk and KPI, then rank the list before funding a build.
Whether the documents and data your first use cases depend on are known, owned and classified, with permissions that an AI answer can respect.
If it scores low: Map the sources for the first two use cases, note who may read each one, and decide which data classes may leave your network.
Whether you have decided where AI may run for each class of data, and whether compute, identity and logging are ready for it.
If it scores low: Size the models your first use cases need, decide the hosting route per data class, and connect AI applications to your identity provider and log pipeline from day one.
Whether policy, inventory, incident handling and audit evidence exist before AI systems reach customers, staff decisions or regulators.
If it scores low: Adopt a governance policy, start a risk register from your AI inventory, and agree who responds when an AI system misbehaves.
Whether staff know how to use AI safely, a team can build and maintain it, and business teams can try ideas without a long queue.
If it scores low: Give every role a short literacy course, name the team that owns AI applications, and let business teams build supervised pilots on an approved platform.
Whether AI in production is evaluated, costed, version-controlled and supervised, so you can tell what is working and what it costs.
If it scores low: Set a test set and a review rhythm for each production use case, record its cost per month, and document every model in a model card.
Each statement scores 0 to 3. The overall score is the share of points earned across all 20, and it places the organisation on one of four levels. Treat the level as the summary and the area scores as the plan.
AI use is informal and mostly individual. Start with a policy, an inventory of what people already use, and one well-scoped pilot with a baseline.
Pilots exist, but they depend on a few people and little is measured. Rank the use cases, fix data access for the best two, and put governance in place before scaling.
Several use cases run with owners and controls. The work now is a shared platform, repeatable evaluation, cost tracking and an audit trail that holds up to review.
Close one gap at a time, starting with the one that blocks your next use case.
Find the lowest area that stands between you and your first or next use case. A low people score matters less than a low data score if the use case cannot reach its documents.
A governance gap becomes an approved policy and a risk register; a data gap becomes a source and permission map for two use cases; an infrastructure gap becomes a hosting decision per data class.
Each deliverable needs one accountable person and a deadline, not a committee. Put it on the same plan as the use case it unblocks.
Ask the same people again once the first use case reaches production, and compare the six area scores with the first run.
An AI readiness assessment checks whether an organisation has what it needs to put AI into real use: clear goals and ranked use cases, data that is known and permissioned, somewhere suitable to run models, policy and audit controls, people who can use and maintain the systems, and a way to measure results. It gives a score per area so you can see which gap will slow you down first. This one takes about five minutes and covers twenty statements in six areas.
Readiness asks whether you can start or scale a specific step, such as moving a pilot into production. Maturity describes how established AI already is across the organisation, usually as levels. The two use the same evidence, so this assessment reports both: a score for each of the six areas, which shows readiness for the next step, and an overall level from Exploring to Operating, which is the maturity view.
Each statement is answered Not yet, Partly, Mostly or Fully, scored 0 to 3. An area score is the points earned as a share of the points available in that area. The overall score is the share across all twenty statements, so areas with four statements weigh slightly more than areas with three. Scores from 0 to 24 read as Exploring, 25 to 49 as Experimenting, 50 to 74 as Scaling and 75 to 100 as Operating.
No. The score is calculated in your browser and the answers are not saved or sent. Site analytics records only that a result was shown and the overall level, without the answers. If you want the editable templates mentioned in the next steps, you can request the template pack by email, which is a separate and optional step.
Fix the lowest-scoring area that blocks your first use case, not every gap at once. Low governance scores usually mean writing a policy and starting a risk register; low data scores mean mapping sources and permissions for one or two use cases; low infrastructure scores mean deciding where models may run for each class of data. Then rerun the assessment after the pilot, with the same people answering, and compare the area scores.
VDF AI installs on your servers, in a private cloud or air-gapped, with permission-aware retrieval, human approval steps and audit trails. An on-premises deployment can start with a three-month trial of the platform.