The Status That Sounds Like Governance but Isn't
You are looking at Halstead & Cole's AI system register. One entry — AIMS-SYS-003, Microsoft 365 Copilot in Outlook — has been marked "Under Review" since the Gap 3 assessment. The label is correct. What does not exist is the procedure that says what "Under Review" actually means. Without it, the label is a placeholder dressed as oversight.
The IT Lead, Maya Okonkwo, opened REG-AIMS-SYS-001 to check the status of AIMS-SYS-003. The entry read: "Microsoft 365 Copilot in Outlook — Under Review — flagged for further assessment during Gap 3 baseline."
She looked for the procedure governing the Under Review status. There was none. Specifically:
No defined clock. How long should a system stay Under Review before a decision is required? No answer in the policy or the procedure file.
No defined reviewer. Who actually owns the reassessment — the IT Lead, the COLP, the AGL, or the Managing Partner? No RACI cell points to the task.
No defined outcome set. What are the legal options at the end of the review? Return to Active with conditions? Move to Suspended pending G8 implementation? Decommission? No document lists the menu.
90 days have passed since the status was assigned. Nobody has raised this as a problem — because the procedure that would have raised it does not exist.
"Without a procedure that defines what each stage means, who decides the transition, and against what criteria, the register's status labels are cosmetic — not governance."
— UNUS London · Gap 5 Article · Cl. 8.2 AnalysisThe gap that turns a risk register into wallpaper
ISO/IEC 42001:2023 requires an AI system to be governed across its entire operational life — from the decision to procure through to the decision to decommission. Without a lifecycle procedure with defined entry and exit criteria at each stage, every other AIMS control loses its operational meaning. The risk register from Gap 4 has scores; this module is the procedure that makes those scores consequential. Gap 5 is rated High because the register's accuracy degrades silently over time — firms rarely notice until an audit, an incident, or a regulatory request forces the question.
What Happens When AI Systems Are Adopted but Never Governed
A missing lifecycle procedure is not an administrative oversight. It is a slow-burn governance failure. Below are the four consequence categories that follow directly from GAP-05 being unaddressed. None of them announce themselves; all of them are visible to an auditor.
Untracked Vendor Updates
OpenAI, Microsoft, and LEAP Legal all update their AI models — sometimes without notice. A model update can change a tool's accuracy, its tendency to hallucinate, or how it handles edge cases. Without a lifecycle procedure that defines a "material vendor change" as a trigger event requiring review, these updates pass through the firm's AI estate undetected. The firm continues using an AI system whose behaviour has changed since the original risk assessment was completed — without realising it.
Scope Creep Without Trigger
An AI tool approved for litigation chronology work quietly gets used in immigration advice drafting. The fee earner treating the tool as a general-purpose assistant has not flagged a scope change — because the policy does not require it, and the lifecycle procedure does not define scope expansion as a trigger event. The risk register now understates the actual deployment. The original impact assessment is no longer relevant to the use case. Governance has drifted.
Ungoverned Decommissioning
When a firm stops using an AI tool, what happens to the data it processed? Without a documented decommissioning procedure: data may not be deleted or retrieved from vendor systems; the system may remain in the register as "Active" when it is not; vendor agreements may auto-renew for another year; and the firm has no audit trail of the decision to retire the tool. A regulator asking "when did you stop using X?" has no answer to point to.
Pemberton Q4 Unanswered
Pemberton Capital's fourth question asks specifically: "how does your firm monitor AI systems after deployment?" The honest answer today is "we don't, systematically." A response of "we review AI use regularly" is not an answer — it is a hope. Without a documented lifecycle procedure, the firm cannot produce a defined monitoring cadence, named trigger events, or an audit trail of reviews. The £620,000 panel relationship is exposed.
The ITIL perspective: lifecycle procedures are operational practice, not paperwork
In ITIL 4 terms, Gap 5 represents a missing Change Enablement and Service Configuration Management practice. An AI system that was approved 18 months ago but has not been re-assessed since is, in ITIL terms, an unconfigured service asset with no defined change records. Lifecycle governance is the procedure that turns AI systems from "things the firm uses" into "services the firm manages." Without it, every other AIMS control operates on an outdated picture of the firm's actual AI estate.
The four non-conformances Clause 8.2 surfaces
Auditors reviewing against ISO 42001 Clause 8.2 typically identify four recurring non-conformities in law firms: NF1 — no lifecycle procedure at all; NF2 — pre-deployment gate exists, post-deployment governance absent; NF3 — vendor model updates not treated as lifecycle events; NF4 — decommissioning not governed. Halstead & Cole currently exhibits NF1 and NF3. PROC-AIMS-LIFE-001 addresses all four.
Reading Clause 8.2 in Plain English
ISO/IEC 42001:2023 Clause 8.2 is short in its written text but wide in its implications. Per ISO/IEC Directives Part 2, the word shall denotes a requirement; should denotes a recommendation; may denotes a permission. The interactive table below lets you filter by obligation type — try it.
Where this clause lives in the standard
Clause 8.2 sits inside Clause 8 (Operation). It is the operative clause for AI systems specifically — distinct from Clause 8.1 (operational planning and control) which deals with general operations. The standard separates lifecycle governance for AI from lifecycle governance for general services because AI systems have characteristics — opacity, vendor dependency, post-deployment behaviour change — that require dedicated procedural treatment.
Risk register
AGL approval
Oversight
Quarterly review
Reassessment
Decision required
90-day max
Board notified
Register closed
Retained 3yr
Why "use" is itself a lifecycle stage — and not an absence of one
Many firms treat "Active" as the end of the governance process — the system is approved, it is in use, governance is complete. ISO 42001 Clause 8.2 treats Active as a stage within the lifecycle, with its own monitoring obligations, quarterly review requirements, and trigger events that can move the system out of Active. A system that has been Active for 18 months with no governance activity is not a governed system — it is an ungoverned system that happens to have been approved at one point in the past. The procedure does not end at deployment; it intensifies.
The supplementary obligation from Clause 6.1
Clause 8.2 closes with the requirement that lifecycle processes "implement the actions determined in Clause 6.1" — the risk assessment and risk treatment actions. This is the integration point with Gap 4: every entry in REG-AIMS-RISK-001 has a residual risk score and a remediation plan. PROC-AIMS-LIFE-001 is the procedure that ensures those scores and plans are actually re-checked at the right cadence. Lifecycle and risk register are co-dependent — without one, the other is academic.
How the SRA's "Ongoing Review" Language Maps to the Standard
The SRA does not use the word "lifecycle." It does not need to. Three existing SRA instruments — applied together — create the same requirement through the language of ongoing supervision. The timeline below maps each instrument to the procedural element of PROC-AIMS-LIFE-001 that satisfies it.
"The use of AI does not diminish a solicitor's personal responsibility for the work produced. Ongoing review of AI outputs is expected, not a one-time assessment."
— SRA · Compliance Tips for Solicitors Regarding the Use of AI and Technology · 9 February 2026SRA Compliance Tips for Solicitors (Feb 2026) · "Ongoing review"
"Ongoing review" is the SRA's plain-English version of the Active stage monitoring obligation in Clause 8.2. A firm that approved GPT-4o in January 2026 but has not reviewed that approval since is not supervising its AI use — it is relying on a historical judgement that may no longer reflect the current tool or the current state of the firm's practice. PROC-AIMS-LIFE-001's quarterly review and eight trigger events are how Halstead & Cole evidences "ongoing review" to the SRA.
SRA Code of Conduct for Solicitors 2019 · Rules 3.2 & 3.3
Rule 3.2 requires solicitors to maintain the level of competence and legal knowledge needed to practise effectively. Rule 3.3 requires supervision of staff and adequate training. The SRA has confirmed both obligations extend to the AI tools used in practice — including understanding their current limitations. Lifecycle governance is how the firm evidences that the competence and supervision obligations are exercised over time, not only at the point of adoption.
SRA Code of Conduct for Firms 2019 · Rule 8.1(a)
The COLP must take all reasonable steps to ensure the firm, its managers, employees, and interest holders comply with their obligations under the SRA's regulatory arrangements. An AI system that has not been reviewed in 18 months cannot be defended as "compliant" under Rule 8.1(a) — the COLP has no evidence on which to base that defence. PROC-AIMS-LIFE-001 provides the documented, dated evidence trail that satisfies this duty.
Ayinde v The London Borough of Tower Hamlets [2025] EWHC 1383 (Admin)
In Ayinde, a litigant relied on AI-generated case citations in a judicial review claim. Several of the citations did not exist. The court drew explicit attention to the responsibility of legal representatives to verify AI-generated material before relying on it. Lifecycle governance — particularly the Active stage monitoring obligation and the T2 incident trigger — is the procedural mechanism through which the firm evidences that verification, and that any verification failure triggers procedural escalation.
One procedure. Two regulators. One questionnaire question answered.
ISO 42001 Clause 8.2 and the SRA's "ongoing review" expectation both require the same thing: a documented, procedurally governed approach to every AI system across its entire operational life. PROC-AIMS-LIFE-001 — built in the next two sections — is the single procedure that satisfies the international standard, evidences compliance with the SRA's supervisory duty, and answers Pemberton Capital's question four with a concrete monitoring cadence rather than a hopeful promise.
The Five Mandatory Stages of PROC-AIMS-LIFE-001
A lifecycle procedure that satisfies Clause 8.2 must define five distinct stages, each with its own entry criteria, stage obligations, and exit criteria. Click each stage below to reveal its content. Together they form the firm's complete lifecycle procedure — the operational engine that makes the risk register consequential.
A system enters Pre-Deploy when a fee earner, practice group head, or IT Lead identifies a candidate AI tool. The entry criteria are concrete: completion of Part A of REG-AIMS-SYS-001 (system description, vendor, data categories, intended scope), completion of Part B of REG-AIMS-RISK-001 (impact assessment with risk tier classification), and the assignment of a named AI System Owner.
The approval gate is not a committee — it is a documented decision with a signature and a date. The AGL (AI Governance Lead) signs the approval. For High-risk systems, the Managing Partner countersigns. No system may enter Active status without passing this gate. The gate is the procedure's first defence against ungoverned deployment.
Entry criteria (all required):
- Part A completed in REG-AIMS-SYS-001 — vendor, model, data categories, intended scope
- Part B impact assessment completed in REG-AIMS-RISK-001 — risk tier assigned (Low / Medium / High)
- AI System Owner named — the individual accountable for the system day-to-day
- Disclosure determination made — does this system's use require client disclosure under G7?
Exit criteria (all required before Active status):
- AGL sign-off on the Pre-Deploy approval gate form
- Managing Partner countersignature (High-risk systems only)
- Register entry updated — status moves from Pre-Deploy to Active with date stamp
An Active system has three standing obligations that operate continuously. Together they convert the risk register from a static document into a living governance instrument — and they are what the SRA means by "ongoing review."
Standing obligations during Active status:
- Monthly model version check by the IT Lead — vendor release notes reviewed and recorded in REG-AIMS-SYS-001. Any model version change = automatic trigger to Under Review
- Quarterly output quality review by the COLP or supervising partner — recorded in the quarterly AIMS performance report. Pattern of errors or near-misses triggers Under Review
- Immediate incident reporting under PROC-AIMS-HITL-001 — any AI incident or near-miss attributable to the system triggers Under Review without delay
Eight defined trigger events (covered in the next section) move a system from Active to Under Review automatically — the IT Lead does not need to make a judgement call; the trigger is the judgement. This is what makes the procedure operable rather than discretionary.
Why Active is not "the end" of governance: Clause 8.2 explicitly requires that use be governed, not just deployment. An Active system is one of the most heavily governed stages in the lifecycle — three continuous monitoring obligations plus eight trigger events. A firm that thinks Active means "no further oversight needed" has misunderstood the clause.
Under Review is the stage that AIMS-SYS-003 has been stuck in for 90 days. With PROC-AIMS-LIFE-001 in place, that status has structure. The stage has a defined clock, a defined assessor, and a defined menu of outcomes — none of which exist in the firm's current arrangement.
The clock: Under Review has a 30-day clock from the trigger event date. The clock starts on the day the system is moved to Under Review and is recorded in REG-AIMS-SYS-001. If the 30-day window expires without a documented decision, the system automatically transitions to Suspended — ambiguity defaults to caution.
The assessor: The AGL leads the reassessment, drawing on the relevant risk register rows and the original impact assessment from Part B of REG-AIMS-RISK-001. The COLP provides input. For High-risk systems, the Managing Partner is consulted before any decision.
The outcome menu (one of three required):
- Return to Active — with updated risk register and, if required, new or strengthened controls. Recorded with a return-to-Active date and updated residual risk scores.
- Transition to Suspended — system removed from client work pending remediation. Triggers the 90-day Suspended clock (Stage 04).
- Decommission — system will not return to Active. Triggers the Decommissioning procedure (Stage 05).
The Under Review decision is recorded on the Under Review Decision Record form (template in PROC-AIMS-LIFE-001), signed by the AGL and (for High-risk systems) counter-signed by the COLP.
A Suspended system may not be used in any client matter. This is a hard rule — it is not a soft preference, and a breach of this rule is reportable to the COLP under the firm's incident procedure. The suspension removes the tool from active deployment while remediation work happens.
Immediate actions on entering Suspended status:
- COLP notifies all fee earners within 24 hours — by email and, where practicable, by team meeting. Notification includes the system name, the trigger reason, and the date the suspension took effect.
- Board (or equivalent senior leadership) is notified within 5 working days. The notification is recorded as a board minute with the suspension reason, the affected matters, and the proposed remediation timeline.
- Register entry updated in REG-AIMS-SYS-001 — status moves to Suspended with date stamp and the 90-day clock start date.
The 90-day rule: The 90-day Suspended clock runs from the date of suspension. If the remediation control is not verified within 90 days, the system is automatically decommissioned. The procedure does not permit indefinite suspension — it forces a resolution.
Returning to Active: A Suspended system may return to Active only when (a) the remediation control is implemented and tested, (b) the residual risk score in REG-AIMS-RISK-001 is updated, and (c) the AGL signs the Return-to-Active form. Without all three, the system remains Suspended and the 90-day clock continues to run.
Decommissioning is a terminal state — the system is permanently retired. But "retired" does not mean "forgotten." Decommissioning triggers three mandatory actions that create an audit trail and protect the firm from residual data exposure.
Mandatory actions on entering Decommissioned status:
- Data deletion or retrieval from vendor systems — documented with written confirmation from the vendor. The confirmation includes the data categories deleted, the deletion date, and the vendor signatory. For systems that processed client personal data, this satisfies the firm's UK GDPR data deletion obligations.
- Vendor agreement review — confirm termination, cancel any auto-renewal clause, and document any data return or transition terms. Many vendor agreements auto-renew on a rolling 12-month basis; missing the cancellation window can result in another year of unnecessary data exposure.
- Register entry closed in REG-AIMS-SYS-001 — decommission date and reason recorded. The entry is retained (not deleted) for three years as the audit trail for the system's entire governance history. The register entry is never deleted — it is the evidence that the system existed, was governed, and was properly retired.
Why Decommissioning is a governed stage, not an absence of one: Firms that treat decommissioning as informal — "we just stopped using it" — leave themselves exposed on three fronts: data protection (UK GDPR), contractual (auto-renewing vendor agreements), and audit (no evidence trail). The Decommissioning checklist within PROC-AIMS-LIFE-001 ensures all three are addressed before the system is closed out.
Important — the procedure is a template, not a final document
The five stages above are the minimum structure required to satisfy Clause 8.2. Firms must adapt each stage to reflect the AI systems they actually deploy and the firm's own approval authorities. A firm with a single COLP may collapse the AGL function into the COLP role; a firm with multiple practice areas may add a practice-area approval step. The procedure is a starting framework, not a finished product. Review by a qualified solicitor or compliance professional is required before adoption.
The Eight Trigger Events That Move a System to Under Review
A lifecycle procedure is only as strong as its trigger events — the defined conditions that automatically move an AI system from Active to Under Review. PROC-AIMS-LIFE-001 defines eight. Each has a named detector, a defined response timescale, and a link back to the part of the AIMS that responds. The table below is the working standard expected by ISO 42001 auditors.
| Trigger | Description | Who detects | Time to trigger |
|---|---|---|---|
| T1 | Model version update Vendor releases a new model version or updates the underlying model without a version change. The AI system has materially changed in behaviour even if the version number has not moved. |
IT Lead Monthly release-note check |
Within 5 working days |
| T2 | Incident or near-miss Any AI incident or near-miss attributable to the system, as defined in POL-AIMS-001 and PROC-AIMS-HITL-001. Includes fabricated citations, incorrect advice, data leakage, and unauthorised disclosure. |
Any fee earner Obligation to report |
Immediately |
| T3 | Scope expansion request Proposal to use the system in a new practice area, matter type, or data category not covered by the original impact assessment. Use outside the registered scope is itself a governance breach. |
Requesting fee earner Or practice group head |
Before expansion |
| T4 | Vendor terms change Material change to vendor data-processing terms, privacy policy, or service agreement. Examples: change to sub-processor list, change to data-retention period, change to model training opt-out. |
IT Lead Quarterly contract review |
Within 5 working days |
| T5 | Regulatory development New SRA guidance, legislation, or court ruling materially affecting the firm's AI use in the relevant context. Includes judicial decisions such as Ayinde [2025] EWHC 1383. |
COLP Regulatory monitoring |
Within 10 working days |
| T6 | Quarterly review finding Quarterly output quality review identifies a pattern of errors, near-misses, or control failures requiring reassessment. One-off issues are handled through supervision; patterns trigger Under Review. |
COLP Or supervising partner |
Within 5 working days |
| T7 | Risk acceptance expiry A formal risk acceptance from REG-AIMS-RISK-001 Section 4 expires without the remediation control being verified. The clock on risk acceptance is part of the register, not the calendar. |
AGL Calendar review |
On expiry date |
| T8 | Supply chain change Material change to the third-party AI supply chain for a self-hosted or vendor-aggregated system. Cross-references Gap 9 (supplier assessment). Includes sub-processor changes and underlying model-provider changes. |
IT Lead Sub-processor monitoring |
Within 5 working days |
The three triggers most firms miss — and why they matter most
The triggers that firms most commonly lack any mechanism to detect are T1 (vendor model update), T3 (scope expansion), and T2 (incident). T1 matters because vendor updates happen frequently and silently. T3 matters because fee earners treat AI tools as general-purpose and use them in unapproved contexts without flagging. T2 matters because firms handle AI errors as client-service issues rather than governance events. Without triggers T1, T2, and T3 detected reliably, the firm's Active stage is ungoverned in practice — regardless of what the register says.
The integration point — lifecycle procedure and risk register are co-dependent
PROC-AIMS-LIFE-001 draws its decision criteria from REG-AIMS-RISK-001 at every gate. The Pre-Deploy gate checks the risk tier. The Under Review reassessment re-scores the affected risk factors. The Suspended-to-Active return requires updated residual risk scores. This co-dependency is intentional: the lifecycle procedure operationalises the risk register — it is the process that makes the register's scores consequential rather than academic. Without one, the other is a document the auditor reads and finds technically present but operationally inert.
How PROC-AIMS-LIFE-001 Is Filed, Signed & Applied to Existing Systems
A lifecycle procedure without document control is a memo. ISO 42001 Clause 7.5 requires that all AIMS documentation carry a defined set of metadata: ID, version, owner, classification, issue date, review cycle, retention, and related documents. The block below shows what Halstead & Cole's control block looks like when PROC-AIMS-LIFE-001 is ready to issue.
PROC-AIMS-LIFE-001 becomes ACTIVE
Status changes from DRAFT to ACTIVE. The date of issue is recorded. The seven-year retention clock starts. The Under Review 30-day clock for AIMS-SYS-003 begins on this date — finally giving the stuck status a defined deadline.
Existing systems applied retrospectively
AIMS-SYS-001, -002, and -004 move from their current informal Active state to a documented Active state. The Pre-Deploy gate form is completed retrospectively with a note that the system was already in use at AIMS implementation date, the risk register scores are attached, and the AGL confirms continued Active status subject to the remediation timeline in the risk acceptance record. AIMS-SYS-003 starts its 30-day Under Review clock on the issue date.
The first Under Review decision is made and recorded
The AGL completes the Under Review Decision Record for AIMS-SYS-003. The decision is one of three: Return to Active with conditions (with updated controls from PROC-AIMS-HITL-001); Transition to Suspended pending full G8 implementation; or Decommission. The decision is signed by the AGL and counter-signed by the COLP.
Question 4 answered with documented procedure
Halstead & Cole responds to Pemberton Capital's fourth questionnaire question with a copy of PROC-AIMS-LIFE-001 attached, the current register status of each system, and the date of the most recent Under Review decision. The answer is no longer "we review AI use regularly" — it is "here is our documented lifecycle procedure, here are the eight trigger events, here is our monitoring cadence."
First quarterly review cycle and T1–T8 monitoring
Monthly model version checks by the IT Lead begin. The first quarterly output quality review is scheduled. The eight trigger events are referenced in the IT Lead's monthly checklist and the COLP's quarterly review template. From this point forward, the Active stage is governed continuously rather than left to drift.
What Gap 5 closure does not, by itself, deliver
PROC-AIMS-LIFE-001 is the operational engine. It does not by itself satisfy the firm's data governance requirements (Gap 6 — covered in Module 07), the client disclosure obligation (Gap 7), or the human-oversight procedure (Gap 8). Each of those documents is a separate procedure that plugs into the lifecycle at specific stages: data governance defines what data may enter Pre-Deploy; disclosure determines whether the system triggers the disclosure obligation at Active; human oversight defines the output-review mechanism that operates throughout Active. Module 07 covers Gaps 6 through 9 in sequence.
Module 06 Completion Checklist
Tick every box below before proceeding to Module 07. This module is the operational engine of the AIMS — if any item is unclear, return to the relevant section before moving on. Each item maps to a specific Clause or SRA obligation that audit will test.
Module 07 — Preview
Module 07 closes the remaining operational gaps. Halstead & Cole will produce its data governance register (REG-AIMS-DATA-001) for Gap 6, its client disclosure procedure (DOC-AIMS-DISC-001) for Gap 7, its human-oversight procedure (PROC-AIMS-HITL-001) for Gap 8, and its supplier assessment regime (REG-AIMS-SUP-001) for Gap 9. With these in place — and the lifecycle procedure from this module governing how each is applied and updated — the AIMS becomes a complete, operational governance architecture rather than a collection of standalone documents.