This document captures the substance of the Project Catalyst program kickoff held August 17, 2026. It is the narrative companion to the Kickoff Presentation Deck — what was presented, what was discussed, and what was decided. It serves as the Day 1 reference for all 262 program participants: who we are, what we are building, how we are governed, and what happens in the first 12 weeks (Phase 0).
1. Opening & Program Authorization
Executive Sponsor M. Kavanagh (ACME COO) opened the kickoff by confirming that the Program Charter was signed August 3, 2026, formally authorizing Project Catalyst as a $99 million, 36.5-month enterprise AI transformation program. The Charter grants Program Director C. Tyrrell (Pulaski Advisory Group) the authority to mobilize 262 people across 25 functional teams, execute three AI capability deployments (BRD-01, BRD-02, BRD-03), and establish an enterprise AI Governance & Center of Excellence.
Kavanagh emphasized one point above all: "We are building AI that is governed from Day 1 — not building AI first and adding governance later. The governance structure, the independent validation function, and the fairness testing standards are not overhead costs to be minimized. They are the reason this program will survive regulatory scrutiny that our competitors will not."
2. Why This Program, Why Now
Two forces make this program both necessary and time-bound:
- Regulatory mandate: CMS-0057-F requires FHIR-based Prior Authorization APIs by January 2027 and imposes decision timelines of 72 hours (urgent) / 7 days (standard). ACME's current manual PA process averages 9.3 days for standard requests — 12% of urgent PAs exceed the 72-hour window. The baseline compliance project handles the API endpoints; Project Catalyst builds the AI capability that processes PA requests at the speed and quality the regulation demands.
- Competitive urgency: UnitedHealth, Humana, and Oscar Health are already running production AI across claims, member engagement, and underwriting. ACME's competitive position erodes with every quarter of delay. Moving fast without governance is risky; moving slowly while competitors deploy is unviable. This program proves the two are not in tension.
3. What We Are Building
| Workstream | Capability | Timeline | Lead |
|---|---|---|---|
| BRD-01 | Claims & Prior Authorization AI — agentic PA automation with mandatory human review for adverse determinations | Year 1 (Aug 2026–Sep 2027) | F. Bennett |
| BRD-02 | Member & Provider Experience AI — conversational virtual assistant + call-center agent-assist | Year 2 (Oct 2027–Sep 2028) | X. Garcia |
| BRD-03 | Underwriting & Risk AI — predictive risk-scoring with actuarial integration and fairness testing | Year 2 (Oct 2027–Dec 2028) | Z. Thompson |
| AI Governance & CoE | Enterprise AI standards, model lifecycle governance, shadow AI prevention, regulatory tracking | Years 1–3 (cross-cutting) | S. Khurana |
| Data & Cloud Platform | Shared cloud platform, MLOps pipelines, data governance, FinOps controls | Years 1–3 (cross-cutting) | W. Kumar |
4. How We Are Governed
Three governance boards operate with concurrent sign-off authority at every phase gate. No single board can override another — if they disagree, the conflict resolution process escalates through joint session, compromise proposal, and ultimately Executive Sponsor resolution.
- Executive Steering Board (M. Kavanagh, 9 members) — business viability, budget, risk appetite, strategic alignment
- AI Governance Board (S. Khurana, 4 members) — AI model risk, responsible AI, fairness/bias, model lifecycle. Has non-overridable blocking authority for model production releases.
- Enterprise Architecture Review Board (D. Chen, 3 members) — technical architecture, security, integration standards
The governance framework is detailed in the Program Governance Model and the AI Governance & JAD Charter. The key structural innovation is the two-line-of-defense model risk structure: the AI CoE (first line, embedded in delivery, advisory) and Independent Model Validation (second line, separate from delivery, blocking authority). No model reaches production without P. Okafor's team signing off — and that sign-off cannot be overridden by schedule pressure or executive direction.
5. Program Director Authority & Operating Model
C. Tyrrell (Pulaski Advisory Group) operates as Program Director under delegated authority from the Executive Sponsor. Key operating parameters:
- Schedule management within ±2 weeks without board approval
- Resource changes affecting fewer than 5 FTEs without board approval
- Procurement decisions under $500K (with CFO concurrence)
- Contingency drawdowns up to $500K (with CFO joint approval and documented root cause)
- Cannot authorize scope changes of any size — all scope changes require board approval
- Cannot override model validation decisions or architecture decisions — those boards own those domains
The operating model is hybrid: phase-gated governance (waterfall-style gates for regulatory and executive sign-off) with agile sprint-level execution within each BRD delivery team (2-week sprints, daily standups, sprint reviews, retrospectives). This is deliberate — regulated healthcare AI requires documented phase gates for compliance, while AI/ML development requires the iterative feedback loops that only agile execution provides.
6. Phase 0: What Happens in the First 12 Weeks
Phase 0 (August 17 – November 6, 2026) is the foundation phase. Nothing is built yet; everything is prepared. The 12 weeks deliver:
- AI Readiness Assessment — baseline measurement of ACME's data quality, infrastructure, team capabilities, and organizational readiness for AI deployment
- AI Governance Framework v1 — CoE Charter approved, NIST AI RMF mapping complete, enterprise AI standards draft published, Independent Model Validation protocol defined
- Platform Vendor Selection — hyperscaler cloud provider(s) selected; BAAs signed; development environments provisioned. Multi-cloud architecture adopted per CN-003.
- Data Profiling (BRD-01) — legacy claims data quality assessed across all 4 source systems. Results determine whether R-01 contingency ($1.2M, pre-approved via CN-001) needs activation.
- Team Mobilization — all Phase 0/1 team members onboarded, compliance briefing completed, system access provisioned, and JAD session schedule published.
- BRD-01 JAD Series Launch — JAD sessions begin in October with all 8 mandatory attendees (see JAD Charter §7). 8 sessions planned through December.
7. Year 1 Key Milestones
| Milestone | Target Date | Gate |
|---|---|---|
| Phase 0 Gate: Foundation & Readiness | 06 Nov 2026 | All 3 Boards |
| BRD-01 Requirements Sign-off (JAD Complete) | 18 Dec 2026 | AI Gov Board |
| Phase 1 Gate: Design Approved | 28 Feb 2027 | All 3 Boards |
| Platform Foundation Go-Live | 31 Mar 2027 | EARB |
| Phase 2 Gate: Build & Test | 30 Jun 2027 | All 3 Boards |
| BRD-01 Pilot Go-Live | 15 Jul 2027 | AI Gov Board |
| Phase 3 Gate: Production Readiness | 31 Aug 2027 | All 3 Boards |
| BRD-01 Full Production | 30 Sep 2027 | ESB |
8. The Ask: What Every Stakeholder Needs to Do
- Executive Steering Board members: Attend bi-weekly board meetings. Review phase-gate packages when distributed. Make decisions within the defined turnaround times. Champion the program internally.
- 25 Team Leads: Attend weekly Program Status Sync (Fridays, 60 min). Update your RAIDD Log entries weekly. Ensure your team members complete onboarding (compliance briefing, system access) within 2 weeks of assignment.
- JAD Mandatory Attendees (8 roles): Block your calendars for the BRD-01 JAD sessions (October–December 2026, typically 2–3 hours per session). Designate a qualified proxy for sessions you cannot attend — notify C. Tyrrell at least 48 hours before any absence.
- All team members: Complete the mandatory compliance & data-handling briefing before accessing any program systems. Offshore personnel: complete the additional data-residency briefing.
9. Where to Find Everything
All program artifacts are accessible from the AI Transformation Suite landing page. Key documents for Day 1:
- Program Charter — what we are authorized to do
- Governance Model — how decisions are made
- Program Management Plan — how we execute
- Resource Plan — who is on the team (262 people, all named)
- Organization Chart — who reports to whom
- RAIDD Log — what risks we are tracking
- Communications Plan — how information flows