Program Charter — Project Catalyst

This Charter formally authorizes the existence of Project Catalyst and grants the Program Director authority to apply organizational resources to program activities. It is the single highest-authority document in the program — all other program artifacts (PMP, Governance Model, BRDs, SOWs, RAIDD Log) derive their authority from this Charter and must be consistent with it. Approved by Executive Sponsor M. Kavanagh (ACME COO) on August 3, 2026. Updates require formal change control with Executive Steering Board approval.

1. Charter Authorization & Document Purpose

This Charter formally authorizes Project Catalyst, a $99,000,000, 36.5-month AI Transformation Program at ACME Highland Health. It grants C. Tyrrell (Program Director, Pulaski Advisory Group) the authority to apply organizational resources — budget, personnel, vendor contracts, and infrastructure — to program activities within the parameters defined herein.

1.1 What This Charter Authorizes

• Expenditure of up to $99,000,000 across three Statements of Work (SOW-01, SOW-02, SOW-03) over 36.5 months
• Mobilization of up to 262 people across 25 functional teams (staffing ramps by phase; not concurrent)
• Execution of three AI capability deployments (BRD-01, BRD-02, BRD-03) plus two cross-cutting workstreams
• Establishment of an AI Governance & Center of Excellence with enterprise-wide governance authority over AI model risk
• Engagement with external vendors (hyperscaler cloud providers, LLM/AI model providers, SaaS tooling vendors) through procurement processes defined in the Program Management Plan

1.2 What This Charter Does Not Authorize

• This Charter does not authorize ACME's baseline CMS-0057-F compliance project, which is a separate initiative with its own charter, budget, and timeline. BRD-01 builds on top of baseline compliance infrastructure; it does not replace it.
• This Charter does not authorize changes to ACME's corporate organizational structure. The AI Governance & CoE operates as a program function during the 36.5-month lifecycle; its transition to permanent ACME operations is a closeout deliverable, not a charter commitment.
• This Charter does not authorize expenditure beyond $99,000,000 without ACME Board-level approval.

1.3 Charter Authority Hierarchy

This Charter is the highest-authority program document. In the event of conflict between this Charter and any subsidiary document (PMP, Governance Model, BRDs, SOWs), this Charter prevails until both documents are reconciled through formal change control. The hierarchy:

1. Program Charter (this document) — authorizes the program
2. Program Governance Model — defines how decisions are made within the authority granted by this Charter
3. Program Management Plan — defines how the program is executed within the governance framework
4. Statements of Work (SOW-01, SOW-02, SOW-03) — contractual scope and fee schedules with Pulaski Advisory Group
5. Business Requirements Documents (BRD-01, BRD-02, BRD-03) — detailed specifications for each delivery workstream
6. All other program artifacts — derive authority from the documents above

2. Strategic Business Case & Market Context

2.1 The Competitive & Regulatory Moment

ACME Highland Health is a national health insurer with approximately 4 million members, operating in a market where AI deployment is no longer a strategic option — it is a competitive requirement. The business case for Project Catalyst rests on two converging forces that make the program both necessary and time-bound:

Regulatory Driver: CMS-0057-F

The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) mandates that impacted payers implement FHIR-based Prior Authorization APIs by January 1, 2027, deliver PA decisions within 72 hours (urgent) / 7 calendar days (standard) effective January 1, 2026, and publicly report PA metrics including approval rate, denial rate, and average decision time. ACME's current manual PA process frequently exceeds these decision windows (average 9.3 calendar days for standard PAs; 12% of urgent PAs exceed 72 hours). While ACME's separate baseline compliance project handles minimum API and reporting requirements, Project Catalyst authorizes the larger, durable capability: production AI at scale that processes PA requests within the compliant infrastructure with the speed, consistency, and transparency the regulation demands.

Competitive Driver: Market AI Adoption

ACME's primary competitors are already deploying production AI across claims, prior authorization, and member engagement:

• UnitedHealth Group: Invested $3.5B+ in AI/technology platforms through Optum; deployed AI-assisted prior authorization and claims processing at scale. Reduced PA turnaround to under 24 hours for standard requests.
• Humana: Deployed conversational AI for member engagement; AI-assisted clinical decision support integrated into care management workflows.
• Oscar Health: Built-from-scratch technology-first insurer; AI-driven member experience and claims processing from inception.

ACME's competitive position erodes with every quarter of delayed AI deployment. Members, employers, and brokers increasingly compare payer performance on digital experience, decision speed, and transparency — all areas where AI provides measurable advantage. Waiting is unviable.

2.2 The Governance-First Strategy

The industry norm for AI deployment is "move fast, add governance later." The results are visible: hallucination incidents (Moffatt v. Air Canada, 2024), regulatory scrutiny of AI-generated claim denials (multiple state investigations, 2024–2026), and reputational damage from bias in underwriting models. Project Catalyst takes the opposite approach: stand up governance before any model hits production. The AI Governance & Center of Excellence, the Independent Model Validation function, and the three-board governance structure are all established during Phase 0 — before BRD-01 begins development. This approach is slower to first delivery than ungoverned AI deployment, but it produces AI systems that are defensible when (not if) regulators, courts, or media scrutinize them.

2.3 Financial Summary

The program's financial case is documented in detail in the Cost-Benefit Analysis and Total Cost of Ownership. Summary:

3. Program Vision & Value Proposition

3.1 Vision Statement

Project Catalyst will establish ACME Highland Health as a governance-first AI leader in health insurance — deploying production AI capabilities across claims processing, member engagement, and underwriting that are accurate, fair, explainable, and defensible, while transitioning AI governance from a program function to a permanent organizational capability.

3.2 Value Proposition

The program delivers value across four dimensions:

• Operational efficiency: PA cycle time reduced from 9.3 days to ≤3 days (standard); call-center handle time reduced 30%; underwriting cycle time reduced from 14–21 days to 7–10 days (large group). These are measurable, auditable outcomes tracked against Phase 0 baselines.
• Risk reduction: AI governance framework, independent model validation, and fairness testing reduce regulatory risk, legal liability, and reputational exposure. The cost of not governing AI (regulatory fines, litigation, member trust erosion) exceeds the cost of governing it.
• Competitive positioning: ACME moves from lagging to leading on AI deployment among national health insurers. Speed-to-decision, digital member experience, and underwriting accuracy become competitive advantages, not liabilities.
• Organizational capability: At closeout, ACME has a permanent AI Governance & CoE function, trained internal teams, and operational AI systems — not a dependency on consulting partners. The program builds the muscle, not just the product.

4. Program Objectives & Success Criteria

Program objectives are defined using SMART criteria (Specific, Measurable, Achievable, Relevant, Time-bound). Each objective has a named measurement owner, a baseline established during Phase 0, and a target verified at program closeout.

5. Program Scope

5.1 In-Scope — Five Delivery Workstreams

5.1.1 BRD-01: Claims & Prior Authorization AI (Year 1 Flagship)

Agentic prior-authorization automation using NLP document extraction, rules-based clinical decision support, and ML prediction to process PA requests with mandatory human review for all adverse determinations. Includes FHIR-based PA API integration (CMS-0057-F compliance), explainable determination generation, fraud pattern detection, and production monitoring with retraining triggers. 28-person delivery team. Full specification in BRD-01.

5.1.2 BRD-02: Member & Provider Experience AI (Year 2)

Conversational AI for member self-service (24/7 virtual assistant across web, mobile, and voice channels) and call-center agent-assist (real-time response suggestions, case summarization). RAG-based factual grounding architecture to prevent hallucination (Moffatt v. Air Canada mitigation). Confidence-based escalation to human agents. WCAG 2.1 AA accessibility. 22-person delivery team. Full specification in BRD-02.

5.1.3 BRD-03: Underwriting & Risk AI (Year 2)

Predictive risk-scoring models that augment actuarial judgment in group and individual underwriting. Explainable gradient boosted ensemble (GBE) models with SHAP-based factor attribution. Mandatory fairness testing with 3% disparate-impact threshold (strictest in program). Proxy variable audit. Shadow-mode pilot. NAIC Model AI Bulletin alignment. Actuarial ASOP compliance. 14-person delivery team. Full specification in BRD-03.

5.1.4 AI Governance & Center of Excellence (Cross-Cutting, Years 1–3)

Enterprise-wide first line of defense for AI model risk. Establishes AI development standards aligned to NIST AI RMF 1.0 and ISO/IEC 42001. Operates the model lifecycle governance process: concept review at JAD sessions, interim reviews during development, coordination with Independent Model Validation, and production monitoring standards. Manages shadow-AI prevention. 12-person team. Transitions to permanent ACME operations at closeout.

5.1.5 Data & Cloud AI Platform Foundation (Cross-Cutting, Years 1–3)

Shared hyperscaler cloud platform, MLOps pipelines, data governance layer, and FinOps controls. Built during Year 1 (Platform Foundation Go-Live target: March 2027) and used by all three BRDs. Multi-cloud architecture (Change Control CN-003). Data residency controls ensuring PHI remains onshore US. 20-person team.

5.2 Explicitly Out of Scope

• ACME's baseline CMS-0057-F compliance project: Minimum API endpoints, decision-timeline enforcement, and reporting mechanics are delivered by a separate project with its own charter and budget. BRD-01 builds on top of this baseline.
• Non-AI claims processing system upgrades: If ACME's core claims system requires modernization to support BRD-01 integration, that modernization is a separate project. BRD-01 integrates with the claims system as-is (or as-upgraded by a separate effort).
• ACME corporate organizational redesign: The AI CoE is a program function; its permanent placement in ACME's org structure is a closeout recommendation, not a charter commitment.
• Provider-facing AI tools beyond BRD-02: Provider-specific AI capabilities (e.g., AI-assisted clinical documentation, provider performance scoring) are out of scope. BRD-02's provider features are limited to provider-facing self-service (eligibility, claims status, PA status).
• Pharmacy benefit management AI: Pharmacy PA and formulary management are handled by ACME's PBM partner. AI integration with PBM workflows is out of scope unless specifically added through change control.

6. Multi-Year Phasing & BRD Structure

6.1 Three-Year Program Structure

6.2 Phase-Gate Structure (Per BRD Cycle)

Each BRD follows a five-phase gated lifecycle. Detailed phase definitions, gate criteria, and sign-off procedures are in the Program Governance Model Sections 7–9 and Program Management Plan Section 4.

7. Key Deliverables & Milestones

8. Governance Structure & Decision Rights

Program governance operates through a four-level authority hierarchy (Strategic → Governance → Program → Execution) with three standing governance boards that hold concurrent sign-off authority at every phase gate. Full governance detail is in the Program Governance Model; this section establishes the governance structure's charter-level authority.

8.1 Three Governance Boards

8.2 Concurrent Sign-Off Requirement

All three boards must sign off independently at every phase gate. One board cannot override another. If boards disagree, the conflict resolution process (defined in Governance Model Section 6) escalates through joint session → compromise proposal → Executive Sponsor resolution. This concurrent structure prevents any single perspective (business, AI risk, or technical) from dominating decisions that affect all three.

9. AI Governance & Model Risk Framework

9.1 Two-Line-of-Defense Model Risk Structure

AI model risk is managed through a deliberate two-line-of-defense structure, modeled on financial services regulatory practice (OCC SR 11-7) and adapted for healthcare AI:

• First Line — AI Governance & CoE (S. Khurana, 12 people): Embedded in delivery. Sets standards, reviews model design at JAD sessions, conducts interim reviews during development, defines fairness and explainability requirements. Advisory and standards-setting during development.
• Second Line — Independent Model Validation (P. Okafor, 6 people): Separate from delivery. Reports to AI Governance Board, not to BRD leads. Tests every model against five criteria (accuracy, fairness, explainability, security, regulatory compliance). Has formal, non-overridable blocking authority for production releases.

9.2 Regulatory Framework Alignment

The AI governance framework aligns to four regulatory/standards anchors:

• NIST AI RMF 1.0: Four-function framework (Govern, Map, Measure, Manage) mapped to program governance mechanisms. See Governance Model Section 14.
• ISO/IEC 42001: AI management system standard; organizational alignment target for the AI CoE.
• NAIC Model AI Bulletin: Insurance-specific AI governance guidance; directly applicable to BRD-03 (underwriting) and BRD-01 (claims/PA).
• State AI-in-insurance statutes: Colorado SB 21-169, Connecticut SB 1103, and emerging state legislation. Compliance tracked quarterly by the AI CoE.

9.3 Responsible AI Principles (Charter-Level)

These principles are non-negotiable across all program workstreams:

1. Transparency: Every AI-assisted decision must be explainable in human-readable terms.
2. Human oversight: Mandatory human review for all adverse PA determinations (BRD-01), underwriting decisions in edge-case bands (BRD-03), and low-confidence member interactions (BRD-02).
3. Fairness: Models tested for disparate impact before production; ongoing monitoring post-production. No model deployed that produces systematically different outcomes for demographically similar populations without documented, defensible justification.
4. Accountability: Every model decision is logged with full audit trail. Every governance decision is documented with named authority. The question "Who approved this?" always has an answer.
5. Privacy: PHI handled exclusively in onshore US environments. Data used for model training is de-identified per HIPAA Safe Harbor. Privacy impact assessments conducted per BRD.

10. Organizational Structure & Roster

10.1 Roster Summary

Total program roster: 262 people across 25 functional teams. This is total headcount across the 36.5-month program lifecycle, not concurrent headcount — staffing ramps by phase (Phase 0: ~80 active; Year 1 peak: ~120; Year 2 peak: ~180; Year 3: ~100 declining). Complete named roster in the Resource Plan; full reporting structure in the Organization Chart.

10.2 25 Functional Teams

11. Program Director Authority & Limitations

11.1 Authority Granted

The Program Director (C. Tyrrell, Pulaski Advisory Group) is granted the following authority by this Charter:

• Day-to-day delivery coordination across all 25 teams (dotted-line matrix authority)
• Schedule management within ±2 weeks of baselined milestones without board approval
• Resource changes affecting fewer than 5 FTEs without board approval
• Procurement decisions under $500,000 (with CFO concurrence)
• Contingency drawdowns up to $500,000 (with CFO joint approval and documented root cause)
• Cross-BRD technical decisions (via Scrum-of-Scrums coordination)
• Escalation to any governance board or Executive Sponsor at any time

11.2 Authority Limitations

• The Program Director cannot authorize scope changes, regardless of size — all scope changes require board approval
• The Program Director cannot override the AI Governance Board's model validation decisions
• The Program Director cannot override the EARB's architecture decisions
• The Program Director cannot approve budget changes >$500,000 or schedule changes >2 weeks without Executive Steering Board approval
• The Program Director cannot alter the governance structure defined in this Charter or the Governance Model without Executive Steering Board approval
• The Program Director does not have solid-line management authority over team members — each team lead reports to their home organization (ACME department head or Pulaski engagement manager). The Program Director coordinates; he does not manage.

11.3 Program Director Accountability

The Program Director is accountable to the Executive Sponsor (M. Kavanagh) for: on-time delivery of all phase-gate milestones, budget management within authorization, risk management per the RAIDD framework, governance compliance per this Charter and the Governance Model, and stakeholder communication per the Communications Plan. The Program Director reports to the Executive Steering Board bi-weekly and at every phase gate.

12. Stakeholder Identification

13. Budget Authorization

This Charter authorizes total program expenditure of $99,000,000 across three fiscal years.

13.1 Budget by Cost Category

14. SOW Structure & Fee Schedule

Milestone payments are tied to phase-gate deliverables. Payment requires formal deliverable acceptance per the phase-gate sign-off procedure. Full SOW detail in SOW-01, SOW-02, SOW-03.

15. Contingency Reserve & Financial Controls

• Contingency Reserve: $9,000,000 (approximately 10% of total budget). Not discretionary — use requires formal approval with documented root cause. Tiered approval: ≤$500K (Program Director + CFO), $500K–$1M (Executive Steering Board), >$1M (Executive Sponsor + CFO). Low-watermark escalation if reserve drops below $2.0M.
• Monthly financial tracking: Program Finance (A. Rodriguez) reports actual vs. budget by category monthly. Variance >5% triggers investigation; >10% triggers Executive Steering Board escalation.
• Earned value: CPI and SPI calculated monthly. Thresholds: Green (≥0.95), Yellow (0.85–0.94), Red (<0.85). Red triggers recovery plan and board escalation.
• No supplemental funding: This Charter does not authorize expenditure beyond $99,000,000. Additional funding would require ACME Board-level approval — an extraordinary measure not anticipated in the base plan.

16. Benefits Realization Framework

Benefits are tracked against baselines established during Phase 0. Full methodology and projections in the Cost-Benefit Analysis. Summary of expected steady-state annual benefits (Year 5+):

Benefits realization is tracked quarterly beginning 6 months after each BRD's production release. Final benefits-realization audit conducted at program closeout and presented to Executive Steering Board.

17. Regulatory & Compliance Framework

18. High-Level Risk Assessment

Ten risks are identified at program initiation and documented in the RAIDD Log with owner, probability/impact score, mitigation strategy, and trigger threshold. Summary:

19. Assumptions

• ACME member volume and PA case mix remain stable: approximately 4 million members, 800,000 annual PA cases. Material changes (>10%) would require re-baselining model training data and capacity planning.
• State regulatory environment remains stable within current risk envelope. New AI-in-insurance mandates are absorbed through CoE standards updates, not model redesigns (unless mandated).
• Cloud/LLM vendor ecosystem remains viable and accessible through program duration. No major platform discontinuations, pricing shocks, or access restrictions requiring platform migration.
• ACME executive commitment (budget, staffing, governance participation) remains through all three years. Budget re-authorization for Years 2 and 3 proceeds as planned at respective Phase 4 gates.
• ACME's separate CMS-0057-F baseline compliance project delivers FHIR API endpoints on schedule (independent project, dependency DEP-01).
• Historical data (claims, PA, underwriting, call transcripts) is available at sufficient volume and quality for model training (validated Phase 0; R-01 addresses data quality risk).

20. Constraints

• Budget ceiling: $99M total authorization. No supplemental funding mechanism without ACME Board approval.
• PHI data residency: All PHI must remain in onshore US cloud regions. Offshore teams access only synthetic/anonymized data.
• Regulatory timeline: CMS-0057-F effective dates are fixed; ACME's compliance project and BRD-01 must deliver accordingly.
• Model validation independence: Independent Model Validation team cannot report to or be combined with BRD delivery teams. Structural constraint.
• Human-in-the-loop: Mandatory human review for adverse PA determinations (BRD-01), pricing decisions (BRD-03), and low-confidence member interactions (BRD-02). Cannot be disabled.
• Actuarial sign-off: Every pricing decision influenced by BRD-03 models requires credentialed actuary sign-off (ASOP No. 56 compliance).

21. Dependencies

22. Document Control

23. Related Documents

• Program Governance Model — defines decision rights, board composition, phase-gate procedures
• Program Management Plan — defines execution, monitoring, and control procedures
• RAIDD Log — risk, assumption, issue, dependency, and decision register
• Resource Plan — complete 262-person named roster
• Organization Chart — full reporting structure
• RACI Matrix — responsibility assignments
• Cost-Benefit Analysis — financial justification and NPV analysis
• Total Cost of Ownership — 10-year cost comparison
• SOW-01, SOW-02, SOW-03 — contractual scope and fee schedules
• BRD-01, BRD-02, BRD-03 — business requirements specifications
• AI Governance & JAD Charter — CoE mandate and JAD session structure
• Change Control Log — formal change record
• Communications Plan — stakeholder communication cadence

24. Approval Signatures

The undersigned approve this Charter and authorize the commencement of Project Catalyst as defined herein.