Where is AI value stalling in your organisation?
Eleven questions, ten minutes, and a scored readiness snapshot across the three domains that decide whether AI investment becomes outcomes. Answered in your browser. Nothing is sent anywhere until you choose to receive the written summary.
Most organisations invest the ratio upside down.
Boston Consulting Group puts the split at ten percent on algorithms, twenty percent on technology and data, and seventy percent on people and processes. Budgets rarely follow that shape. Technology vendors sell the ten percent because it is what they have to sell, systems integrators staff the twenty percent because it bills predictably, and the seventy percent has no natural vendor, so it goes unowned.
That is why pilots succeed and rollouts stall. The pilot proves the model works. The rollout requires decision rights, an accountable owner, a change path through a union or a bargaining agent, a records and retention answer, and a way to tell a minister or a board what happens when the system is wrong. None of that is a technology problem, and none of it appears in a proof of concept.
For federal departments there is a further clock. The Treasury Board Directive on Automated Decision-Making requires an Algorithmic Impact Assessment before an automated decision system goes into production, with the assessment published. Departments that have not established who owns that assessment discover the gap at the worst possible moment: after the tool is built and before it can be used.
Three domains, eleven questions, weighted 70-20-10.
Each question is scored zero to four, normalised to 100 within its domain, then weighted. The weighting is where the diagnostic differs from a generic maturity model: a strong score on technology cannot carry a weak score on governance, because the weighting will not let it.
People, process and governance
Decision rights, accountable ownership, change capacity, workforce capability, policy and assurance. The domain that decides whether anything reaches production.
Data and technology
Data quality, access, lineage and the platform decisions that either enable the use case or quietly cap it.
Use case and AI
Whether the use case is defined against a real business outcome, and whether the model choice follows from it or precedes it.
A score, a breakdown, and a written summary you can hand to someone.
At the end you see a score out of 100, a tier, and a breakdown by domain. You can stop there. If you want the written executive summary, you provide a work email and it is generated on screen, ready to print or save as a PDF.
The summary is written to interpret your scores. It does not compute them: the scoring runs entirely in your browser before any email is entered. The optional cost-of-delay figure is modelled, not measured, and the diagnostic states the assumption on screen rather than presenting the number as a finding.
It is a triage, not an audit.
Eleven questions cannot audit an organisation, and this does not pretend to. It is an indicative snapshot designed to tell you which domain to look at first, and it is deliberately built so that a sceptical reader can see how every number was produced.
The 10-20-70 formulation is BCG’s. The questions, the scoring and the cost-of-delay model are AltNexus’s, and are not endorsed by BCG. Where the instrument models rather than measures, it labels the figure as modelled.
The paid AI Velocity Gap Diagnostic is a different thing: a fixed-price engagement in three tiers, two to four weeks, that quantifies the cost in your own numbers and returns a sequenced, costed roadmap with an outcome guarantee.
Ten minutes, and you own the output.
Eleven questions. No account required to answer. The score is yours whether or not you ever speak to us.
An indicative self-assessment, not an audit and not professional advice. Where a figure is modelled rather than measured, the instrument says so on screen and shows the assumption.