Delivery lead
Accountable for building, buying and running AI services.
- Often held by
- a chief information officer, chief technology officer or head of digital
- Accountable for
- 18 controls
- Contributes to
- 41 controls
- Relies on
- Business owner (12), Security lead (10), Data owner (9), AI governance lead (5), Risk lead (3), Executive sponsor (2), People lead (2)
- Supports
- AI governance lead (9), Business owner (9), Data owner (6), Executive sponsor (5), Security lead (5), People lead (5), Risk lead (2)
Accountable for
This role answers for each of these controls reaching the level the organisation has set, and holds the evidence.
Safe experimentation
Staff can try AI quickly within defined guardrails covering approved environments, data limits, time limits and exit criteria.
With: Security lead, Data owner, AI governance lead.
Evidence an assessor would expect
- The experimentation rules, covering environments, data, time limits and exit
- A log of recent experiments with their start, end and outcome
Reuse across the organisation
Successful patterns, components and lessons from uses of AI are recorded so others can find them, and are reused before anything new is built.
With: AI governance lead, Business owner.
Evidence an assessor would expect
- The catalogue or library of reusable patterns, components and lessons
- Proposal or gate records showing the reuse check
Build, buy or configure approach
A consistent, documented method decides for each use of AI whether to configure tools already owned, procure, build, or wait, and the reasons are recorded.
With: Business owner, Security lead, AI governance lead.
Evidence an assessor would expect
- The sourcing method or decision guide
- Recorded sourcing decisions for a sample of uses
Running cost management
The ongoing cost of each AI service, including usage-based charges, is known, forecast, assigned to an owner and kept within agreed limits.
With: Business owner, Executive sponsor.
Evidence an assessor would expect
- Cost reports by AI service, with budget holders
- Usage limits, alerts or forecasts in place for usage-priced services
AI-aware procurement and contracts
Contracts for AI products and services set clear terms on data rights, intellectual property, notice of model changes, transparency and exit.
With: Risk lead, Data owner, Security lead.
Evidence an assessor would expect
- Standard AI contract clauses or schedule
- A sample of recent AI contracts showing those terms
Supplier performance and exit
AI suppliers are managed against agreed performance measures, and each critical AI service has a tested route to exit or replace the supplier.
With: Business owner, Data owner.
Evidence an assessor would expect
- Supplier performance reports and review records
- Exit plans for critical AI services, with any test results
Delivery standards for AI
AI work is delivered by multidisciplinary teams using user-centred methods to defined delivery standards.
With: Business owner, People lead.
Evidence an assessor would expect
- The delivery standards or playbook for AI work
- Team composition and user research records for a sample of projects
- Accessibility test records from before go-live
Service management for live AI
Every live AI service has defined support, change and problem management arrangements, including handling of changes to models and prompts.
With: Business owner, Security lead.
Evidence an assessor would expect
- Service catalogue or support records for live AI services
- Change records covering model or prompt updates
Continuity when AI fails
Services that depend on AI can continue safely when the AI is unavailable, degraded or producing wrong results, through planned fallbacks that have been tested.
With: Business owner, Security lead.
Evidence an assessor would expect
- Impact analysis of AI unavailability with recovery objectives for each dependent service
- Continuity plans or service designs showing AI dependencies and fallbacks
- Fallback test records with results and actions
Approved AI tools and platforms
The organisation keeps a current list of approved AI tools and platforms, each with clear conditions of use, and staff know where to find it.
With: Security lead, Data owner, AI governance lead.
Evidence an assessor would expect
- The approved tools list with conditions of use, owners and review date
- The request process and a log of recent requests and decisions
- Approved build, test and run environments with owners
Reference architecture for AI
Agreed architecture patterns set out how AI components fit into the technology estate, and new uses of AI follow them or record why not.
With: Security lead, Data owner.
Evidence an assessor would expect
- The reference architecture or pattern library for AI
- Design review records showing patterns applied or departures approved
- Output marking in generative AI patterns, where applicable
Hosting and data location
Requirements for where AI is hosted and where its data is processed and stored are defined, and each use of AI is checked against them before and during live use.
With: Security lead, Data owner, Risk lead.
Evidence an assessor would expect
- Documented hosting and data location requirements
- Location checks for a sample of AI services, including subprocessor information
Integration with existing systems
AI is connected to core systems through supported, documented interfaces that the organisation can maintain, change and switch off.
With: Data owner, Business owner.
Evidence an assessor would expect
- Integration documentation showing interfaces, data flows and owners
- Support arrangements for AI integrations
- Permission and action limits for AI systems and agents, with approvals and risk assessments
Managing models in live use
Models, prompts and their configuration are versioned and released under control, so the organisation knows what is running and can roll back to a known good version.
With: Business owner, Security lead.
Evidence an assessor would expect
- Version history for models, prompts and configuration
- Release records with approvals
- A record of a rollback or rollback test
- Documentation for the current version, written for users, operators and regulators
- Use documentation stating purpose, users, limitations and where not to rely on it; advance notices of changes
Testing and evaluation before release
Before release, each AI system is tested for accuracy, robustness and behaviour against acceptance criteria agreed in advance, and released only if it meets them.
With: Business owner, Data owner, Risk lead.
Evidence an assessor would expect
- Acceptance criteria and test plans for a sample of releases
- Test results and release sign-off records
- Review records for the testing approach, and harmful output tests for generative AI
- Defined operating conditions and their enforcement
Technology change and refresh
The organisation plans for changes to models, suppliers and platforms, including suppliers withdrawing models, so that AI services continue without disruption.
With: Business owner, Executive sponsor.
Evidence an assessor would expect
- Bill of materials for each AI service with its last review date
- A dependency record showing models, platforms, versions and known end dates
- Refresh plans and budget lines for planned change
- Dependency records showing concentration on single models or suppliers
Unapproved AI tools
Use of unapproved AI tools is detected, and people using them are steered to an approved alternative or a route to get their need met.
With: Security lead, AI governance lead, People lead.
Evidence an assessor would expect
- Detection reports or records of unapproved AI tool use
- Records of responses, including tools added to the approved list as a result
Monitoring live AI
Live AI is monitored against its measures for accuracy, drift, misuse and unexpected behaviour, with thresholds that trigger action.
With: Business owner, Data owner.
Evidence an assessor would expect
- Agent action logs, and a tested way to stop each agent
- Monitoring dashboards or reports for live uses of AI
- Records of threshold breaches and the action taken
- Event logs for live uses of AI, with their retention period
- Monitoring against operating conditions
- Live fairness and sensitive data monitoring results
Contributes to
Another role is accountable; this one does part of the work or supplies what it needs.
Investment and funding for AI
The organisation has a defined approach to funding AI, covering what is paid for centrally and locally, how long commitments last, and how requests for money are judged.
Accountable: Executive sponsor.
Horizon scanning and external expectations
The organisation tracks changes in AI technology, regulation, peer practice and public expectation, and feeds what it finds into its AI strategy.
Accountable: AI governance lead.
Multi-year AI roadmap tied to planning
The organisation has a phased AI roadmap that is linked to its business and financial planning cycles and shows the points where it will decide whether to proceed.
Accountable: Executive sponsor.
Named owners for each use of AI
Every use of AI has named owners for the business outcome, the data, the technology and its assurance, recorded where others can find them.
Accountable: AI governance lead.
AI inventory
The organisation keeps a current inventory of the AI it uses, including AI built into bought products and services, with enough detail to govern it.
Accountable: AI governance lead.
Fairness and bias management
Unfair bias is identified, tested for and reduced at each stage of a use of AI, from design through to live operation.
Accountable: Business owner.
Human oversight proportionate to risk
Each use of AI has human oversight designed to match its risk, stating who reviews, at which point, and with what authority to override or stop it.
Accountable: Business owner.
Risk of AI from suppliers
Before AI in a bought product or service is used, and when it changes, the organisation assesses what it does, how it uses data, how the supplier changes the model, and what assurance the supplier offers.
Accountable: Risk lead.
Idea intake open to all staff
Anyone in the organisation can propose a use of AI through a simple, well-known route, and every proposal is acknowledged and considered.
Accountable: AI governance lead.
Agreed prioritisation criteria
Proposed uses of AI are ranked against published criteria covering value, feasibility, risk, cost and strategic fit, applied the same way to every proposal.
Accountable: AI governance lead.
Pipeline register of proposed uses
A single register records every proposed and in-flight use of AI with its status, owner and decision history, and hands each entry to the AI inventory when it goes live.
Accountable: AI governance lead.
Stage gates from idea to retirement
Each use of AI passes through defined stages (explore, pilot, scale, operate, retire), with entry and exit criteria checked at each gate.
Accountable: AI governance lead.
Proportionate business case
Each use of AI has a business case scaled to its size and risk that states expected benefits, full costs and assumptions, and is approved before significant commitment.
Accountable: Business owner.
Pilot to production decisions
Every pilot ends in a recorded decision to scale, extend or stop, made against criteria and evidence agreed before the pilot started.
Accountable: Business owner.
Retiring uses of AI
Uses of AI that no longer earn their place are withdrawn through a planned process that covers users, dependent processes and disposal or retention of models and data.
Accountable: Business owner.
Target operating model for AI
The organisation has a defined, approved operating model that states who builds, buys, runs and supports AI, and what is done centrally and what locally.
Accountable: Executive sponsor.
Collaboration and sharing
The organisation works with peer organisations and partners to share AI solutions, lessons and capacity where this serves its aims.
Accountable: Executive sponsor.
Involving people in AI design and evaluation
Evaluations and research that involve people have their informed consent, ethics review and protection, use participants who represent those the service will serve, and for significant uses draw on the communities affected.
Accountable: Business owner.
Data strategy supports AI
The organisation's data strategy states what its planned uses of AI need from data and sets out how those needs will be met.
Accountable: Data owner.
Data quality fit for purpose
The quality of data used by each AI system is assessed against what that use needs, and shortfalls are fixed or their effects managed before and during live use.
Accountable: Data owner.
Finding and understanding data
Data available for AI is catalogued and described well enough that people can find it, understand what it means and its limits, and use it correctly.
Accountable: Data owner.
Records management for AI
Retention and records rules cover what AI systems take in and produce, including prompts, outputs and the decisions they supported, and are applied in practice.
Accountable: Data owner.
Controlled access and sharing
AI systems and their users can reach only the data they need for the agreed purpose, and data is shared into and out of AI systems only under agreed rules.
Accountable: Data owner.
Provenance of training and test data
For data used to train, tune or test AI, the organisation knows and records where it came from, what rights it holds over it, and any limits on its use.
Accountable: Data owner.
AI security baseline
The organisation has adopted a recognised AI security baseline, built it into its security standards, and applies it to every use of AI.
Accountable: Security lead.
AI threat modelling
Each use of AI has its AI-specific threats assessed before it goes live and again when it changes, and the results shape its design and controls.
Accountable: Security lead.
Security of the AI supply chain
The security of models, data, code and services obtained from others is assessed before use and managed for as long as they are in use.
Accountable: Security lead.
AI in security incident response
Security incident management covers incidents involving AI, with playbooks, people who have the skills to handle them, and the ability to contain an affected AI service quickly.
Accountable: Security lead.
Security testing of AI before release
Before release, each AI system is security tested against the threats in its threat model by suitably skilled testers, including checks that it does not disclose personal or sensitive information.
Accountable: Security lead.
Specialist roles and careers
The organisation defines the specialist AI roles it needs, with career paths and a professional community that help people grow and stay.
Accountable: People lead.
Skills assessment
The organisation knows which AI skills it has and which it needs, by role and group, and uses this to plan learning, recruitment and workforce decisions.
Accountable: People lead.
Role-based learning
People can get learning matched to what their role needs to do with AI, beyond the baseline for all staff, and use it in their work.
Accountable: People lead.
Bringing in and keeping expertise
The organisation recruits and keeps the AI expertise it needs and uses external experts in a way that transfers knowledge, so that capability stays when contracts end.
Accountable: People lead.
Support for staff using AI
People using AI day to day can get practical help, current guidance and answers to their questions quickly from a known place.
Accountable: People lead.
Planned approach to AI change
AI adoption is managed as organisational change, with a change plan, named sponsors and the resources to carry it out.
Accountable: Executive sponsor.
Learning from what did not work
Experiments and uses of AI that fail or are stopped are reviewed without blame, and the lessons are recorded and shared.
Accountable: AI governance lead.
Adoption support for affected teams
Teams whose work changes because of AI get practical support through the change, and actual adoption is tracked against what was planned.
Accountable: Business owner.
Performance measures and baselines
Each use of AI has agreed performance measures and a recorded baseline of current performance before it goes live.
Accountable: Business owner.
AI incidents and near misses
AI incidents and near misses that are not security incidents, such as wrong outputs, harm to people or misuse, are reported, investigated and learned from.
Accountable: AI governance lead.
User feedback
Feedback from people who use AI and people affected by it is gathered on a regular basis, analysed, and acted on.
Accountable: Business owner.
Serious AI incidents and external notification
The organisation defines serious AI incidents, can suspend a use of AI when needed, and tells suppliers, regulators and affected people through set routes and within set deadlines.
Accountable: Risk lead.