The Tuesday the Questionnaire Arrived
You are standing in Priya Anand's shoes. The largest panel client has just sent over a five-question AI governance questionnaire with thirty days to respond. Question one is the only one that matters right now — and she cannot answer it. This is the gap your module closes.
Priya opened the firm's document management system and searched for "AI policy". No results.
She broadened the search to "artificial intelligence" and "AI governance". A single document appeared: a two-paragraph note from the IT Director dated March 2025, suggesting that fee earners "exercise caution" when using AI tools.
The note had no signatory, no version number, no review date, and no definition of what "caution" means. It could not be presented to Pemberton Capital as evidence of governance. It could not answer question one. And because question one cannot be answered, none of the four questions that follow it can be answered either — without a policy, the firm has no governance mandate under which any control exists.
"Without a policy, there is no governance mandate for any of the controls that the subsequent questions ask about."
— UNUS London · Gap 1 Article · Cl. 5.2 AnalysisThe gap that makes every other gap ungoverned
ISO/IEC 42001:2023 recognises ten control objectives (Annex A.2 through A.10). Eight are technical — registers, risk assessments, lifecycle controls, supplier reviews. The other two are foundational: the AI Policy (A.2) and the assignment of roles (A.3). Until POL-AIMS-001 exists, none of the other eight controls have an authorised mandate to operate. This is why Gap 1 is rated Critical and why it is the first document this programme builds.
What Happens When the Policy Does Not Exist
A missing AI policy is not a paperwork issue. It is an exposure issue — and the exposure touches the COLP personally, not just the firm. Below are the four consequence categories that follow directly from GAP-01 being unaddressed.
Revenue at Risk
Pemberton Capital's panel relationship is worth £620,000 per year. The questionnaire explicitly asks whether the firm has a documented AI policy. A "no" answer likely triggers either contract renegotiation or removal from the panel. This is the most immediate, most visible consequence of GAP-01.
COLP Personally Exposed
Under SRA Code of Conduct for Firms 2019, Rule 8.1(a), the COLP must take "all reasonable steps" to ensure compliance. The SRA's December 2025 thematic review found that only one COLP could outline all their responsibilities. A COLP who has not put AI governance in place cannot evidence "reasonable steps" — that is a personal regulatory risk for Priya Anand.
Fee Earners Acting Without Guidance
Sophie Chen, a junior paralegal in the dispute resolution team, uses GPT-4o for legal research. Without a policy, she has received no formal direction on acceptable use, prohibited use, or the requirement to verify AI-generated research before relying on it. Her work product is the firm's work product — and the firm has no policy framework defining the standard she should meet.
Court Process Exposure
If Sophie submits AI-generated research that contains a fabricated citation — exactly what happened in Ayinde [2025] EWHC 1383 — the firm has no documented policy framework under which the failure was foreseeable, forewarned, or governed. A policy is the firm's first line of evidence that it has discharged its supervision obligation.
The ITIL perspective: this is a P1 governance incident
In ITIL 4 terms, GAP-01 is a Priority 1 incident in the firm's governance service. The AI Policy is the foundational policy document that acts as the ITIL "Policy Management" practice anchor for all AI-related service assets. Without it, no downstream AI governance control has an authorised policy mandate — exactly as ITIL requires all service management practices to operate within an approved policy framework. Halstead & Cole's governance operates on goodwill. Goodwill is not a control.
Real consequence — illustrative precedent
Similar AI-governance questionnaires are already arriving at firms across the UK legal market. Once one major firm answers "yes" to question one, the standard cascades. K&L Gates' March 2026 ISO 42001 certification explicitly stated that "procurement departments and RFP processes will start asking for it." That future is now.
Reading Clause 5.2 in Plain English
ISO/IEC 42001:2023 Clause 5.2 is short — but every word in it carries weight. 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 5.2 sits inside Clause 5 (Leadership). It is one of only two clauses in ISO 42001 that places a direct obligation on top management — the other being Clause 5.3 (organisational roles). This is deliberate: the standard requires a person of sufficient seniority to be publicly accountable for the AI policy before any technical control can be considered active.
Why "top management" matters — Clause 5.2 is not delegable downward
The obligation to establish the AI policy rests on top management. A policy signed by a compliance officer, the IT Director, or the COLP alone does not satisfy Clause 5.2. The signature authority must be the Managing Partner, the Senior Partner, or the equivalent senior leadership body. This is not a bureaucratic detail — it establishes the governance mandate that authorises every other AIMS control. At Halstead & Cole, the COLP maintains the policy; the Senior Partner signs it.
The supplementary obligation in Annex A.2
Annex A control A.2 supplements Clause 5.2 by requiring that information about AI system objectives, behaviour, and limitations shall be made available to relevant internal and external stakeholders. In a law firm context, this extends the policy obligation to client-facing transparency — covered separately in Gap 7 (DOC-AIMS-DISC-001), but the AI policy itself must acknowledge that the transparency obligation exists. The complete solution requires both documents.
How the SRA's Existing Rules Reach the Same Answer
The SRA has not issued a separate AI policy mandate. It does not need to. Three existing SRA instruments — applied together — create the same requirement through the Code of Conduct. The mapping below shows you exactly which rules the AI policy satisfies at the same time.
"It should always be made clear to clients where they are interfacing with AI."
— SRA Compliance Tips for Solicitors Regarding the Use of AI and Technology · Updated 9 February 2026SRA Compliance Tips for Solicitors Regarding the Use of AI (Feb 2026)
The SRA has issued formal compliance guidance on AI use. The three operative obligations are: solicitors must understand the AI tools they use; clients must always be told when they are interfacing with AI; solicitors remain personally responsible for work produced with AI assistance. This guidance was last updated on 9 February 2026 and is the SRA's current position.
SRA Code of Conduct for Solicitors 2019 · Rules 3.2 & 3.3
Rule 3.2: a solicitor must maintain the level of competence and legal knowledge needed to practise effectively. Rule 3.3: a solicitor must supervise their staff and ensure adequate training. The SRA has confirmed that these obligations extend to the AI tools used in practice — which means a solicitor using an LLM for legal research must understand the tool's limitations, including its tendency to produce plausible but false citations.
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. Without a documented AI policy, the COLP cannot evidence having taken reasonable steps to govern AI use. This is the direct line of personal exposure for Priya Anand.
SRA Compliance Officers: A Thematic Review (Dec 2025)
The SRA's own December 2025 review found that only one COLP audited could outline all their responsibilities. The review explicitly flagged that COLPs need to understand obligations arising from technology decisions made by fee earners — not just decisions made by management. An absent AI policy is, today, an obvious flag for any future SRA thematic review of COLP effectiveness.
One document. Four obligations. Two regulators.
ISO 42001 Clause 5.2 and SRA Rules 3.2 / 3.3 / 8.1(a) all require the same thing: a documented, leadership-approved, communicated statement of principle on the firm's use of AI. POL-AIMS-001 — built in the next section — is the single document that satisfies both the international standard and the SRA Code simultaneously. One signature, two regulators covered.
The Six Mandatory Sections of POL-AIMS-001
A policy that satisfies Clause 5.2 must contain six specific elements. Click each section title to reveal the content Halstead & Cole will include. Together they form the firm's complete AI policy — and the foundation for every document in modules 03 to 10.
This opening section establishes the firm's overall stance. It satisfies the Clause 5.2 requirement that the policy "include a commitment to satisfy applicable requirements" and "include a commitment to continual improvement."
POL-AIMS-001 — Section 1 (drafted text):
"[Firm name] shall use AI systems in the delivery of legal services only in a manner that is responsible, transparent, and auditable. The firm shall maintain appropriate human oversight of all AI-generated work product used in legal services. The firm is committed to the continual improvement of its AI governance framework, and to satisfying all applicable requirements set out in ISO/IEC 42001:2023, the SRA Code of Conduct, UK GDPR, and any other regulatory instrument that applies to the firm's use of AI."
This section defines, in plain language, what fee earners may do with AI tools. It satisfies the Clause 5.2 "should" recommendation that the policy defines acceptable use cases in terms staff can apply in practice. Each use case listed below is conditional on the human oversight requirements in Section 4.
Permitted uses include:
- Legal research assistance — where every output is independently verified by a qualified solicitor before reliance
- Document drafting assistance — where all AI-generated content is reviewed and edited by the responsible fee earner
- Document summarisation for internal review purposes only (not for client delivery without rewrite)
- Administrative tasks including billing narrative drafting, scheduling, and internal correspondence
- Client communication drafting — where outputs are reviewed by the responsible fee earner before sending, and the AI Disclosure Statement is referenced where relevant
The list must be adapted to reflect the actual tools Halstead & Cole uses. A firm using only Microsoft Copilot will have a narrower list than a firm deploying bespoke API-connected LLMs.
Equally important as defining what is permitted is defining what is forbidden. This section satisfies the Clause 5.2 expectation that the policy address limitations clearly and operationally.
Prohibited uses include:
- Submitting AI-generated legal research to a court, tribunal, or regulatory body without independent verification by a qualified solicitor
- Using AI to conduct client due diligence, AML risk assessment, or GDPR compliance activities as a substitute for qualified human assessment
- Uploading client documents or client personal data to any AI system not listed in the firm's AI System Register (REG-AIMS-SYS-001) and approved under the firm's data governance controls
- Providing autonomous legal advice to clients via an AI system without qualified solicitor review and sign-off
- Using personal or consumer-grade AI accounts (for example, free-tier ChatGPT) for any matter-related work — the firm's approved enterprise tools and accounts only
These prohibitions are not suggestions. A breach of this section is a breach of the firm's policy and is reportable to the COLP under the firm's incident procedure.
Human oversight is the single most important operational safeguard. This section translates the policy principle into a binding operational requirement, satisfying ISO 42001 Clause 9.1 (monitoring and measurement) and SRA Rule 3.3 (supervision).
POL-AIMS-001 — Section 4 (drafted text):
"Human oversight of AI outputs is a non-negotiable requirement of this policy. No AI-generated output shall be submitted to a client, filed with a court or tribunal, or used as the sole basis for a legal decision without review and sign-off by the responsible qualified solicitor. The sign-off shall be documented in the firm's matter management system."
The detailed procedural requirements — minimum review by output type, sampling cadence, sign-off record format — are defined in PROC-AIMS-HITL-001, the Human-in-the-Loop Oversight Procedure, covered in Module 05. POL-AIMS-001 establishes the principle; the procedure defines the mechanics.
Transparency is the firm's most visible AI obligation under the SRA's February 2026 guidance. This section commits the firm to a defined disclosure practice, satisfying both the SRA's "make clear to clients" requirement and Annex A.2 (transparency).
POL-AIMS-001 — Section 5 (drafted text):
"The firm shall make clear to clients where AI systems are involved in the delivery of legal services, in accordance with the SRA's AI guidance (February 2026) and the firm's AI Transparency Disclosure Statement (DOC-AIMS-DISC-001). The disclosure shall be provided at matter opening and whenever a materially new AI system is introduced to the delivery of the client's work."
The full disclosure wording and the trigger criteria for re-disclosure are covered in Gap 7 (DOC-AIMS-DISC-001), addressed in Module 04. POL-AIMS-001 acknowledges the obligation; the disclosure document operationalises it.
Clause 5.2 requires "a commitment to continual improvement of the AIMS." This section defines what continual improvement looks like operationally: an annual review, three defined review triggers, and a named owner.
POL-AIMS-001 — Section 6 (drafted text):
"This policy shall be reviewed annually as a minimum by the COLP, with the outcome of the review presented to senior leadership for approval. An immediate review shall be triggered by: a material change to applicable regulatory requirements; the deployment of a significant new AI system; or any AI-related incident classified as High or Critical. The COLP shall maintain the policy as a controlled document in the firm's document management system."
Without a defined review cycle, a policy becomes a historic artefact. The "annual minimum with immediate trigger" formulation is the working standard expected by ISO 42001 auditors.
Important — the policy is a template, not a final document
The wording above is the minimum content required to satisfy Clause 5.2. Firms must adapt Section 2 (acceptable use) and Section 3 (prohibited use) to reflect the AI systems they actually deploy. A firm using only Microsoft 365 Copilot has a narrower acceptable-use list than a firm using API-connected LLMs and bespoke document automation. The template is a starting point, not a finished document. Review by a qualified solicitor or compliance professional is required before adoption.
How POL-AIMS-001 Is Signed, Version-Controlled & Filed
A policy without document control is not a controlled document — it 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 POL-AIMS-001 is ready to sign.
POL-AIMS-001 becomes ACTIVE
Status changes from DRAFT to ACTIVE. The date of issue is recorded. The seven-year retention clock starts. Every other AIMS document now has an authorised policy framework under which to exist — Module 03 through 10 documents all cite POL-AIMS-001 in their standard references.
Every fee earner receives and acknowledges the policy
Clause 5.2 requires the policy to be communicated within the organisation. Best practice: a 30-minute briefing session per practice area, followed by an acknowledgement log. Each fee earner signs to confirm they have read Sections 2, 3, and 4 — the operationally relevant sections. The COLP retains the log.
Question 1 is answered "yes" with confidence
Halstead & Cole now responds to Pemberton Capital's questionnaire with a copy of POL-AIMS-001 attached. The £620,000 panel relationship is protected. The first of five questions is answered — and the firm's broader AIMS work has a defensible foundation in place.
Modules 03 to 10 build the remaining eight documents
GAP-01 is closed. Gaps 2 through 9 remain — roles, AI system register, risk assessment, lifecycle controls, data governance, transparency, human oversight, and supply chain. Each is the focus of one subsequent module. All of them cite POL-AIMS-001 as their policy authority.
What Gap 1 closure does not, by itself, deliver
POL-AIMS-001 is the foundation. It does not by itself satisfy the firm's risk assessment obligation (Gap 4), the data governance requirements of UK GDPR (Gap 6), or the supplier assessment regime for the firm's unknown vendor (Gap 9). Each subsequent module builds one of those documents. The order matters — eight of the nine subsequent documents require a signed AI Policy to cite. This module goes first for that reason.
Module 02 Completion Checklist
Tick every box below before proceeding to Module 03. This module is the foundation — if any item is unclear, return to the relevant section before moving on. Each item maps to a specific Clause or SRA Rule that audit will test.
Module 03 — Preview
Module 03 moves from policy to people. Halstead & Cole will produce its governance structure: the RACI Matrix (RACI-AIMS-001), assigning the AI Risk Owner, AI System Owners, AIMS Programme Lead, and Human-in-the-Loop Reviewers. Once these roles exist with named individuals, the firm will be able to answer questions 3 and 4 of Pemberton Capital's questionnaire — and complete the operational half of the AIMS foundation.