Senior Responsible Officers (SROs) and boards are increasingly asked to approve digital and AI‑enabled investments on the basis of strategic cases and target outcomes. The Green Book (2026) update (published 28 October 2026) tightens appraisal expectations: economic value must now be demonstrably linked to deliverability. That changes the question a board must answer. The decision at stake is not only “Should we fund this?” but “Have we been given procurement‑ready evidence that the change is executable, reversible and auditable?”.
The executive decision: fund procurement and contract award, or require further delivery evidence
Boards routinely face a binary question dressed as a continuum: approve funding to proceed to procurement/award; or refuse until delivery evidence meets a standard. For digital and AI work, the right answer depends on whether the strategic case has been translated into procurement‑grade artefacts that make acceptance testable, controllable and reversible. The specific decision to document is: authorise transition from funded discovery/design to a procurement or delivery contract, subject to conditional acceptance evidence and defined cutover gates.
Why this problem is persistent: appraisal divorced from executable evidence
Green Book updates emphasise better, proportionate appraisal for government interventions. Yet discovery and business cases still too often stop at ‘required functionality’ and high‑level benefits. Delivery fails when benefits, risks and costings are not mapped to the practical means of delivery: measurable acceptance criteria; workforce capacity and TUPE impacts; supplier reversibility; data lineage for migrated records; and regulatory constraints such as the Data (Use and Access) Act (2025) and ICO expectations on automated decision‑making. Failure modes include late discovery of incompatible supplier interfaces, untestable reconciliation requirements, unstaffed runbooks, and contract terms that prevent timely reversibility.
How the elements interact: making appraisal operational
Turn the appraisal into an evidence flow where each strategic claim is traceable to an operational artefact. Example: a benefits line (reduction in manual casework) must map to required system changes, process redesign, workforce training, SLOs that measure throughput, an acceptance test that reconciles outputs to source records, and contract clauses obliging suppliers to support reversibility. Data and regulatory constraints (DUAA, UK GDPR) sit across that chain: they influence what data can be migrated, whether operational extracts may be used for model retraining, and which controls are mandatory prior to acceptance.
Suppliers and contracts are not separate levers — they are operational dependencies. A composed acceptance ledger needs supplier‑anchored tests (SLIs/SLOs implemented or verifiable by each supplier), supplier reversibility obligations, escrow or portability clauses, and penalty/remedy triggers that map to the acceptance tests. Cyber controls should be translated into pass/fail gates tied to supplier evidence and live recovery tests; the NCSC Cyber Assessment Framework v4.0 provides the outcome structure that procurement and delivery teams should map to contractual pass/fail conditions.
The relevant stages of the transformation lifecycle (and the artefacts leaders must insist on)
1) Discovery and options appraisal — deliver a procurement‑ready Strategic & Deliverability Case: not just benefits and estimates but an acceptance ledger that maps each major benefit and risk to measurable acceptance tests, owners, datasets and supplier dependencies.
2) Procurement specification — require machine‑readable composed acceptance tests in the call‑off and contract: SLI/SLO definitions, reconciliation tests, data migration snapshots, runbook versioning and escrow triggers. Include supplier reversibility and live recovery test obligations in the contract schedule.
3) Build and integration — mandate traceable requirements with test scripts, test datasets and a composed test runner that can execute end‑to‑end reconciliation and report pass/fail with forensic evidence.
4) Pre‑transition readiness — require staffed runbooks, evidence of workforce availability and TUPE mapping, SRE‑oriented SLO dashboards, and business continuity scenarios exercised with suppliers.
5) Acceptance & cutover — gate approvals on the acceptance ledger passing composed reconciliation tests, verified escrow snapshots, and live supplier reversal drills where practicable.
6) Transition to BAU — require transfer artefacts, operational runbooks, SLIs/SLOs in contracts and an early stabilisation metric that is outcome‑focused (not an arbitrary timebox imposed by procurement).

Concrete artefacts and the minimum acceptance pack for a funding decision
Boards should require the following procurement‑ready artefacts before authorising funding to proceed to award or delivery: 1) A composed Acceptance Ledger: machine‑readable mapping of benefits ➜ acceptance tests ➜ data sources ➜ owners ➜ supplier clauses. 2) Testable SLO catalogue: SLIs, calculation method, collection method, SLO target and error budget policy (OpenSLO or an equivalent machine‑readable format is recommended). 3) Data governance statement: DUAA/UK GDPR impact, permitted uses, migration snapshots, provenance and retention rules. 4) Workforce and TUPE evidence: role maps, staff transfer triggers, required training hours and a staffed rota for the initial operational period. 5) Supplier reversibility and escrow schedule: scope (code, configuration, data, runbooks), automated verification steps and live recovery test obligations. 6) Integration and reconciliation test suite: executable scripts, synthetic and live test data, pass/fail criteria and forensic snapshots.
These artefacts convert appraisal outputs into contractible promises and testable acceptance gates.
Failure modes boards must watch for (and simple mitigations)
Failure: acceptance tests that measure availability but not data integrity. Mitigation: require composed reconciliation tests that exercise migrated records and produce forensic proof (snapshots + checksums + reconciliation report).
Failure: supplier contract contains ambiguous portability or escrow language. Mitigation: require a verifiable escrow process with staged verification triggers and defined deliverables on contract award.
Failure: workforce evidence is a narrative, not testable. Mitigation: require staff lists, contracts/TUPE triggers, shadow shifts and signed acceptance by operational leads that they can staff scenarios in the runbook.
Failure: AI use is declared late and operational data has been used for retraining. Mitigation: demand a data‑use gate that documents whether operational extracts may be used for model training and the governance around it (DPIA, DUAA compliance, ICO guidance).

Decision gates: how to word conditional funding approvals
A template board decision should make approval conditional and explicit. Example wording: “The Board authorises funding to proceed to procurement and contract negotiation on the condition that the programme supplies a procurement‑ready Acceptance Ledger, supplier reversibility and escrow terms accepted by Commercial, a verified SLO catalogue and a signed workforce readiness statement from the Operations Director. If any composed acceptance tests fail at pre‑transition, the SRO must pause cutover and present a remedial plan to the Board within ten working days.”
This makes the Board’s authority active rather than ceremonial: approval is granted subject to testable artefacts and remedial obligations.
What success looks like and how to measure it post‑decision
Success is not a binary ‘go‑live’ but the ability to demonstrate, on demand, that each board claim maps to live evidence: reconciliations that pass, SLO dashboards showing error budgets in range, a verified escrow snapshot, and live runbook rehearsals with supplier reversals executed. Require a small set of post‑award metrics for the early stabilisation period that are outcome‑oriented (for example, reconciliation pass rate, time to reverse to baseline, staff‑to‑demand ratio) and make them contractually reportable.
Boards should insist on an evidence trail that lets them answer: could we reverse this transition within the contractual and operational windows if we needed to?
A board that funds delivery without procurement‑grade acceptance evidence is buying a promise; a board that insists on an Acceptance Ledger is buying a testable and reversible outcome.
Antares recommended actions
These are Antares's recommended first actions for organisations turning the issues in this article into practical governance and delivery.
- Require a procurement‑ready Acceptance Ledger as the minimum deliverable for any board approval to proceed to contract award. The ledger must map benefits and risks to composed acceptance tests, owners, datasets and contract clauses.
- Mandate a testable SLO catalogue for any regulated or mission‑critical service: SLIs, collection method, SLO targets, error budgets and the automated dashboards that will evidence them (OpenSLO‑compatible where practicable).
- Make funding conditional on verifiable supplier reversibility: explicit escrow/portability schedules, contractual triggers for release and staged live recovery tests before acceptance is granted.
- Include workforce evidence as an explicit gate: staffed runbooks, signed operational ownership of critical functions, and a TUPE mapping with named roles and conditional staffing triggers.
- Require a regulatory compliance gate for data and AI: a DUAA/UK GDPR statement, DPIA where applicable, and AI assurance evidence referenced to the Cabinet Office question bank where automated decision‑making is used.
- Draft conditional board approvals that require remedial plans within tight windows if composed acceptance tests fail; make remedial timelines and escalation explicit in the decision record.
Our insights are provided for general information only and reflect the position at the date of publication. They do not constitute legal, financial, regulatory, cybersecurity or other advice tailored to your circumstances and should not be relied upon as a substitute for appropriate professional advice.
While we take reasonable care over our content, we do not guarantee that it is complete, accurate or current. Reading our insights does not create a client relationship with Antares Consultancy. To the fullest extent permitted by law, we accept no liability for decisions made or losses arising from reliance on this content. External links are provided for convenience and do not imply endorsement.