New: The State of AI Assurance 2026 is out — download it free.
Talk to QapitolRun Snapshot
How It Works · AI Assurance Lifecycle

A continuous loop, because AI doesn’t hold still.

Find, evaluate, control, verify and monitor AI systems continuously — across six stages that run as a loop, not once at launch. Each stage names what Qapitol does, in context, never as a one-time gate.

The assurance lifecycle, running as a loop
CONTINUOUSNEVER ONE-TIMEthe signal never stops travelling06 → 01 · THE LOOP CLOSES01Plan & Govern02Design & Specify03Build & Integrate04Evaluate & Assure05Release & Verify06Run & Optimise
Why a loop, not a launch gate

A loop, not once at launch.

Six stages, running as a loop. Read them top to bottom — then notice the sixth hands straight back to the first. Each stage names what Qapitol does, in context, never as a one-time gate.

AI systems don’t hold still. They drift, they get re-prompted, the data underneath them changes — so the work of standing behind them can’t stop at a single sign-off. It runs as a continuous loop, where the last stage feeds the first.

The six stages

Each stage names what Qapitol does, in context.

Six stages, running as a loop. Read them top to bottom — then notice the sixth hands straight back to the first.

  1. STAGE 01

    Plan & Govern

    Set the obligations, risk classification, and sign-off criteria each AI system must meet — before anything is built.

  2. STAGE 02

    Design & Specify

    Define what “correct” means where there is no single right answer — the evals and controls are specified up front, not retrofitted.

  3. STAGE 03

    Build & Integrate

    Wire assurance into how systems are built, not bolted on after. Controls travel with the system from the start.

  4. STAGE 04

    Evaluate & Assure

    Score behaviour, test failure modes adversarially, and surface what cannot be signed off — across App, Agent and Data Assurance.

  5. STAGE 05

    Release & Verify

    Gate releases on evidence; produce the audit-ready proof that a control was applied, so leadership can sign with confidence.

  6. STAGE 06

    Run & Optimise

    Monitor behaviour and drift in production — because AI systems change after launch — and feed what you learn back into the loop.

06 → 01 · the loop closes
How the loop holds

Run & Optimise monitors behaviour and drift in production — because AI systems change after launch — and feeds what you learn back into the loop. What you find in production reshapes how you Plan & Govern the next turn.

The last stage feeds the first.

Run the loop on your own AI estate.

Start with a Snapshot to see where you stand across all six stages.