Benefits Realization Plan

Executive Summary

The Cost-Benefit Analysis recommended proceeding with Project Catalyst on a conditional basis: a positive NPV of $2,968,904 and an 8.94% IRR that clears the 8% hurdle by a modest margin, explicitly “conditioned on disciplined benefit tracking”. This plan is that condition, made into an instrument. It names who owns each of the four benefit streams, what each reports before any dollar is realized, and how the annual benefits audit — which the Closeout Report already assumes exists — actually runs.

1 · Why this program needs the plan the other four did not

Every suite in this portfolio has a benefits plan, and each makes a different argument. Catalyst’s is the sharpest, because Catalyst’s business case is the only one that is conditional on its own execution. The federal case handed benefits to another party; the PM case handed them to another department; the agile case could actually read them. This case can do none of those things — it has to earn a return that a single slip can erase.

This document supplies the instrument. It does three things the CBA deliberately left to a later artifact: it assigns each benefit stream a single accountable owner; it defines the leading indicators that show whether a benefit is on track to realize, years before it realizes in cash; and it specifies the annual audit that the closeout report already promises but that nothing in the suite had yet defined.

2 · Benefit register

Four streams, totalling $27,200,000 at full run-rate. No stream is invented — each is a row of the CBA benefit model, here given an owner, a leading measure and a link to the RAIDD risk that most threatens it.

3 · What each stream reports before it realizes

Because no benefit produces cash during the three-year build, the plan reports leading indicators throughout. A leading indicator shows whether the capability that produces a benefit is arriving on schedule; it is not the benefit, and the plan never reports it as one. But it is the only thing that can be read in Years 1–3, and reading nothing for three years on a case this exposed would be negligent (BG-03).

One measurement rule governs all four. Underwriting (BEN-04) may not report a risk-selection lift until the model has passed independent validation — the CBA models that benefit “post independent validation” for a reason. Reporting a lift from an unvalidated model is how a benefit case comes to rest on a number the model was never entitled to produce.

4 · The ramp is a commitment, not a forecast

The CBA’s ramp curve is often read as prudence — a conservative estimate of how fast benefits arrive. For benefits realization it is the opposite: it is the schedule the program is committing to hit. Each year’s figure is a checkpoint, and a miss is reported as a miss, not absorbed by sliding the curve to the right.

5 · The annual benefits audit

The Closeout Report states that benefits ownership transfers to ACME’s AI Center of Excellence and Finance, who “continue the benefits audit annually,” with the first audit 90 days after closeout. Until this plan, that audit had no defined scope, owners or cadence — the closeout was pointing at a control that did not exist. Here is the control.

6 · What the numbers depend on — and what happens when they move

The CBA’s sensitivity analysis is not decoration; it is the list of things the audit must watch, because each is a lever on a case with little slack. Restated as benefit-realization exposures:

All four downsides are one-directional or heavily skewed to the downside. That asymmetry is the whole reason benefit tracking is a condition of approval: the case has real upside if execution is clean, but its risks subtract without a matching chance to add. An audit that only reported realized dollars would see these too late; an audit that watches the four leading drivers sees them while there is still time to act.

7 · What the model does not capture

The CBA lists indirect benefits it deliberately excluded from the NPV: a governed AI capability that improves ACME’s regulatory posture ahead of tightening AI-in-insurance oversight; a reusable platform and Center of Excellence that lower the cost of every future AI use case; member and provider experience gains with second-order revenue effects; and durable in-house AI/MLOps capability. These are real and they are correctly outside the model — a benefits plan that quietly monetized them to pad the case would undermine the discipline the whole document exists to enforce.

8 · Open gaps

Recorded with owners rather than left implicit. None is a defect in another artifact except BG-01 and BG-02, which are — and both are fixable in the documents named.

9 · Realization status

10 · Approval