Portfolio governance

The decision is the beginning of the accountability, not the end of it

Approved initiatives carry their conditions, their review dates and the forecast that justified them. Twelve months later the question is whether the promised benefit arrived, and the platform is built so that question has an answer.

From verdict to initiative

Only certain verdicts can become an initiative. An outcome of “use conventional automation” or “further evidence required” does not convert into an AI investment, because it was not one.

When an approved decision converts, it brings its conditions, its required controls, its review date and its expiry with it. The financial forecast is frozen at that moment and cannot be edited afterwards.

The initiative lifecycle. Transitions are enforced, so a stage cannot be skipped to make a report look tidier.
StageWhat it meansCan move to
ProposedConverted from an approved decision, not yet funded.Approved, or cancelled
ApprovedFunded, with conditions and a frozen baseline attached.In delivery, on hold, or cancelled
In deliveryBeing implemented. Spend and milestones are recorded against it.Live, on hold, or cancelled
On holdPaused. Remains visible in the portfolio rather than disappearing from it.In delivery, or cancelled
LiveIn production. Benefit realisation is measured against the baseline.Under review, or retired
Under reviewPost-implementation review in progress.Live, or retired
RetiredWithdrawn from service. The record and its outcome remain.Terminal
CancelledStopped before completion. The reason is recorded rather than deleted.Terminal

The frozen baseline

The most common way an organisation avoids learning from a failed initiative is by quietly revising the forecast until the outcome matches it. DUN-AI makes that impossible.

Sealed at approval

Forecast cost, forecast benefit, the assumptions behind them and the horizon are captured when the decision is sealed, and the record cannot be modified.

Measured against, not alongside

Actual spend and realised benefit are compared to the frozen figures. A revised forecast can be recorded, but it never replaces the baseline that justified the money.

Variance is reported with its direction stated in words as well as sign: spending less than forecast and delivering less benefit than forecast are both variances, and only one of them is good news.

Post-implementation review

A post-implementation report states what was promised, what was delivered, what the difference was, and — the part usually omitted — whether the original decision was sound given what is now known.

What a post-implementation report is required to contain.
SectionRequirement
Forecast against actualBoth figures, the variance, and the currency. No blending across currencies and no midpoint substituted for a range.
Conditions of approvalEach condition, and whether it was met. An unmet condition on a successful initiative is still an unmet condition.
Assumption reviewWhich assumptions held, which did not, and what that implies for similar cases in future.
Decision qualityWhether the decision was reasonable on the evidence available at the time — which is a different question from whether it worked out.
Unrealised benefitStated explicitly where benefit did not materialise, with what is known about why. Not omitted, and not offset against something else.

What the portfolio view shows

  • Initiatives by stage, and how much is actively being spent
  • Approved value, actual spend, forecast benefit and realised benefit, reported per currency and never summed across currencies
  • Overdue reviews, listed by name rather than counted
  • Expired decisions still in delivery — an initiative continuing under an authority that has lapsed
  • Conditions of approval that remain outstanding

How an assessment reaches a decision →