What does an AI-readiness assessment include?
A useful AI-readiness assessment qualifies one valuable workload. It identifies the current failure, accountable owner, approved evidence, data routes, human authority, evaluation cases, technical options, operating burden, and stop conditions. Its output is a decision record—not a generic maturity score or a predetermined hardware quote.
Reviewed 12 October 2026 · No confidential source material is required for initial qualification.
The assessment should make a deployment decision possible.
It must convert an AI idea into a bounded operating case that can be tested, priced, accepted, deferred, or stopped.
Current-state map
Users, steps, systems, delays, failure modes, consequence, demand, and accountable owner.
Source map
Authoritative documents, data classes, quality gaps, permissions, retention, and prohibited flows.
Decision rights
What the system may prepare, recommend, read, write, or never do without human approval.
Acceptance set
Representative cases, expected results, failure classes, thresholds, latency, and recovery tests.
Qualified options
Instructions, retrieval, tools, model routes, deployment lanes, capacity, integrations, and controls.
Go, experiment, hold, or stop
Recommended path, scope, costs to validate, risks, assumptions, exclusions, and no-go conditions.
What the assessment should not pretend to prove.
It is not a compliance certificate.
It identifies material controls and evidence gaps. Legal, privacy, security, and sector obligations still require the appropriate accountable reviewers.
It is not a hardware sales script.
No appliance, memory class, model, or hosting lane should be recommended until a representative workload has been measured.
It is not “AI everywhere.”
Readiness is workload-specific. One good candidate does not authorize broad deployment across unrelated work.
It is not production acceptance.
The assessment defines the test. A later prototype or deployment must pass it with observable evidence.
Choose a workload with a real owner and consequence.
Describe the job, present failure, evidence, users, constraints, and desired result. HONONOH can then define the smallest defensible validation loop.