This report formally closes Project Catalyst, the AI Transformation Program delivered for ACME Highland Health by Pulaski Advisory Group. It records final performance against baseline, benefits realized, deliverable acceptance, outstanding items, team performance, and administrative closure. It is issued at program end following completion of knowledge transfer and a ≥ 4-week period of independent ACME operation of all delivered AI capabilities.
1 · Purpose & Overview
Project Catalyst was chartered to establish enterprise AI capability at ACME Highland Health across three workstreams and a shared platform and governance foundation, over a 36.5-month, $99.0M program. Its purpose was to modernize prior-authorization and claims decisioning, member and provider experience, and underwriting risk assessment — safely, under a formal AI governance model, and with measurable benefit. This report confirms the program met that mandate, documents the final results, and releases the program's resources and contractual obligations. Governance authority transfers to ACME's now-permanent AI Center of Excellence.
2 · Performance Against Baseline
| Dimension | Baseline | Actual at Closeout | Result |
|---|---|---|---|
| Schedule | 36.5 months | 36.5 months | On time · SPI 1.00 |
| Cost | $99,000,000 | $98,400,000 | $0.6M under · CPI 1.01 |
| Contingency reserve | $9,000,000 | $8,400,000 used | $0.6M returned |
| Scope — workstreams | 5 delivered | 5 delivered | Complete |
| Scope — BRD production go-lives | 3 (BRD-01/02/03) | 3 | All in production |
| Major milestones met on time | 15 | 15 | 100% |
The program delivered on its planned end date and closed $0.6M under the authorized budget, having drawn $8.4M of the $9.0M contingency reserve against risks that materialized during delivery (principally legacy claims-data remediation and a mid-program GenAI compute-cost adjustment). No annual SOW envelope was exceeded; two formal re-baselining events occurred within contingency and were approved by the Executive Steering Board without increasing the total authorized budget.
3 · Benefits Realization
Benefits tracking against the approved Cost-Benefit Analysis confirms the program is delivering to plan as it enters steady state:
| Benefit | CBA Target | Realized / Run-Rate | Status |
|---|---|---|---|
| Annual steady-state benefits | $27.2M | $27.2M run-rate | On plan |
| Net Present Value | +$2.97M | Tracking positive | Confirmed |
| Payback period (simple) | ~7.6 years | On trajectory | On plan |
| Prior-auth decision turnaround | Material reduction | Achieved (BRD-01) | Realized |
| Member self-service containment | Improved | Achieved (BRD-02) | Realized |
4 · Deliverables & Acceptance
| Deliverable | Owner Acceptance | Date | Status |
|---|---|---|---|
| BRD-01 · Claims & Prior Authorization AI (production) | ESB / Utilization Management | 30 Sep 2027 | Accepted |
| Data & Cloud AI Platform + MLOps | EARB | 12 Mar 2027 | Accepted |
| AI Governance Framework & CoE (operational) | AI Governance Board | 06 Nov 2026 → | Accepted |
| BRD-02 · Member & Provider Experience AI (production) | ESB / Customer Operations | 30 Sep 2028 | Accepted |
| BRD-03 · Underwriting & Risk AI (production, post shadow-mode) | ESB / Actuarial | 31 Dec 2028 | Accepted |
| Knowledge transfer & steady-state handover | ESB | 30 Jun 2029 | Accepted |
5 · Outstanding Items & Follow-on Actions
| Item | Owner (post-program) | Target |
|---|---|---|
| First annual benefits audit | ACME AI CoE / Finance | Closeout + 90 days |
| Continuous model monitoring & scheduled retraining cadence | ACME AI CoE (MLOps) | Ongoing steady-state |
| Minor enhancement backlog (non-blocking) | ACME product owners | Steady-state roadmap |
| Vendor license renewals & FinOps optimization | ACME Procurement / Finance | Annual |
No critical or high-severity items remain open at closeout. All production capabilities operated independently by ACME for more than four weeks prior to this report.
6 · Team Performance
The program mobilized a life-of-program roster of 262 people across 25 functional teams, ramping by phase from a foundation-heavy Phase 0 to peak concurrent staffing in Year 2 (BRD-02/03) and tapering through the Year 3 transition. The blended ACME-FTE and Pulaski onshore/offshore model held, with onshore-only staffing preserved for PHI claims-decision, Legal, Compliance, and SOX roles throughout. The two-line-of-defense model-risk structure — delivery teams as first line, Independent Model Validation as an independent second line — functioned as designed and was cited by the AI Governance Board as a key control. No key-person risk materialized into a delivery impact; planned substitutions were absorbed without milestone slippage.
7 · Lessons Learned Summary
Fifteen formal lessons were captured across governance, AI/ML delivery, data, financial, and change-and-adoption themes. The most consequential:
- Front-load data quality. Legacy claims-data fragmentation was the program's largest realized risk; early Phase-0 profiling and a funded remediation effort protected the BRD-01 timeline and should be a Phase-0 default on any AI program.
- Budget for AI compute as a variable, not a fixed cost. GenAI compute/token spend required FinOps guardrails and a mid-program adjustment; establishing cost telemetry and alerting before scale-up prevented overrun.
- Human-in-the-loop controls paid for themselves. Designing disclaimer, confidence-threshold, and human-review controls into member-facing AI from the outset (informed by the Air Canada precedent) meant no liability incident occurred.
- Independent validation is worth the queue time. The second-line IMV gate caught a fairness issue before BRD-03 production and forced a retraining cycle — a delay that avoided a far costlier post-production problem.
- Phased, opt-in adoption beat big-bang rollout. Frontline adoption resistance was managed by phased rollout and added change-management investment rather than mandate.
The complete register is maintained in the Lessons Learned Register.
8 · Administrative Closure
| Closure Activity | Status |
|---|---|
| All three annual SOWs delivered and financially settled | Closed |
| Final invoice reconciliation & contingency return ($0.6M) | Complete |
| Vendor contracts transitioned to ACME steady-state ownership | Transferred |
| Program records, model cards, and governance artifacts archived | Archived |
| Access, environments, and credentials transitioned or decommissioned | Complete |
| Consultant staff released; ACME steady-state team confirmed | Complete |
9 · Approval
| Role | Name | Decision | Date |
|---|---|---|---|
| Program Director (Pulaski) | C. Tyrrell | Submitted for closure | 29 Aug 2029 |
| Executive Sponsor (Chair, ESB) | ACME Executive Steering Board | Approved — Program Closed | 29 Aug 2029 |
| AI Governance Board | S. Khurana (Director) | Governance handover accepted | 29 Aug 2029 |
| Program Finance | A. Rodriguez | Financial close confirmed | 29 Aug 2029 |