Stage four · Readiness Roadmap

The Assessment Ends With Owned Action, Not a PDF Score

A readiness result is only useful if it changes what happens next. The roadmap converts discovery and assessment findings into a governed action register.

Every action traces to a question, a gate or an evidence gap. Each one records an owner, a dependency, the evidence required to close it, a decision gate, and a reassessment trigger. Completing an activity is not closure. Evidence accepted by a named validator is.

Roadmap principles

Not a generic list of recommendations.

Every action must trace to evidence from the organisational discovery, the use-case screening, the detailed assessment, the critical gates and evidence gaps, and the obligations and assurance that apply. Priority is set in this order.

  1. Prohibited or legally blocked activity.
  2. Red critical gates.
  3. Fixed regulatory, contractual or operational deadlines.
  4. High-consequence unknowns.
  5. Amber critical gates.
  6. Cross-cutting organisational dependencies.
  7. The lowest detailed-assessment dimensions.
  8. Evidence-confidence gaps.
  9. Value optimisation and scale opportunities.
Workstreams

Twelve workstreams, used only where they apply.

Findings are grouped so ownership and dependencies are clear. No action is forced into a workstream that does not fit it.

  1. W1 Strategy, sponsorship and value
  2. W2 Governance, accountability and risk classification
  3. W3 Process discovery and redesign
  4. W4 Data, knowledge, context and memory
  5. W5 Architecture and integration
  6. W6 Security, identity and permissions
  7. W7 Agent design, authority and human oversight
  8. W8 Legal, regulatory, ethics and affected-person rights
  9. W9 Workforce, roles, skills and change
  10. W10 Testing, simulation and independent assurance
  11. W11 Observability, incident response and resilience
  12. W12 Supplier, economics and sustainability
The action register

What every action records.

Action
A specific outcome to be established, not a vague activity.
Source finding
The assessment question, gate or evidence gap it traces to.
Why it matters
The value, consequence, obligation or dependency behind it.
Accountable owner
One role or person accountable, with the contributors required.
Dependency
The actions or decisions that must precede it.
Required evidence
The artefact or operating evidence that proves completion.
Priority
Critical, High, Medium or Later, from a transparent scoring formula.
Validator
The person or function that accepts the closure evidence.
Residual risk
The remaining exposure after the action is complete.
Reassessment impact
The questions, gates, dimensions and autonomy decision affected.
Example action

Implement a unique non-human identity and task-bounded permissions for the complaint-resolution agent

Source finding
P6.2 red (unique agent identity); P6.3 amber (least-privilege access)
Why it matters
Shared user credentials prevent attribution and exceed the authorised task
Accountable owner
Head of Identity and Access Management
Required evidence
Identity record, approved entitlement design, access test and revocation test
Reassessment impact
P6.2, P6.3, gates G10 and G11, and the autonomy ceiling
Decision gates

Seven points where a governing body decides.

Progress between stages is a decision, not a default. Each gate names the evidence required before it can be passed.

Gate 0 Discovery complete

Scope and multidisciplinary participants agreed; material organisational unknowns recorded; use case and process owner identified; nothing obviously prohibited is being progressed.

Gate 1 Screening passed

Screening context complete; no red stop condition; required detailed-assessment participants identified; principal evidence sources available.

Gate 2 Detailed design and evidence complete

Detailed assessment completed and challenged; red gates closed or the blocked action removed; intended autonomy and human controls documented; legal classification complete; test and assurance plan approved.

Gate 3 Bounded experiment authorised

Acceptance thresholds agreed before testing; representative test environment approved; identity, least privilege, logging and containment operate; human oversight staffed; experiment limits and stop conditions documented.

Gate 4 Controlled production validation

Assessment meets the R4 conditions; required gates green or approved exceptions; end-to-end and recovery testing passed; residual risk formally accepted; monitoring and incident response active.

Gate 5 Governed scaling or authority expansion

Sustained operating evidence at the current authority level; benefits and economics validated; incidents, overrides and differential outcomes reviewed; capacity, resilience and concentration risk reassessed at scale; independent assurance where consequence warrants it.

Gate 6 Retirement or replacement

Retirement trigger or replacement decision documented; access and credentials revoked; data, memory and records retained or deleted appropriately; open actions reconciled; supplier exit and continuity complete; lessons incorporated.

Closure criteria

An action is closed only when.

  • The required outcome exists across the defined scope.
  • The required evidence is attached or referenced.
  • The named validator has accepted the evidence.
  • Residual risk is recorded and owned.
  • The affected assessment answers and gate statuses have been reassessed.
  • The readiness decision and autonomy ceiling have been recalculated if affected.

Publishing a policy or buying a tool is not closure unless the required operating outcome is evidenced.

Reassessment triggers

Readiness is reassessed when.

  • A change in the agent's objective or authorised actions.
  • An increase in autonomy, volume, transaction value or affected population.
  • A new model, major model version or material supplier change.
  • A change to prompts or instructions, tools, integration, permissions, memory or knowledge source.
  • A process, policy, system, data or workforce change.
  • A new or changed legal, regulatory or contractual obligation.
  • An incident, near miss, complaint, material override trend or control failure.
  • Performance, cost or benefit outside approved thresholds.
  • A new agent-to-agent delegation or orchestration.
  • Entry into a new geography, customer group or regulated activity.
  • A scheduled review date.
Roadmap views

One set of actions, four audiences.

Executive view

Decisions, red and amber gates, owners, funding, benefits and dates.

Delivery view

Workstreams, dependencies, milestones and blockers.

Assurance view

Obligations, controls, evidence, validators and residual risk.

Process and agent view

Dimension changes, autonomy ceiling and reassessment history.

After the assessment

Turn the result into a governed plan.

The roadmap is produced as part of the facilitated detailed assessment, or run as a continuing governance workspace. It does not replace accountable approval, specialist advice, regulatory obligations or independent assurance.