← AI Transformation Suite Change Control · Change Request Detail

Change Request CR-003

Download Word

A worked example of the program's change-control process at full depth. The Change Control Log is the register of all changes; this is the underlying document for a single request — the artifact that a Tier 2 change actually requires before any board will consider it. CR-003 was raised in Week 4 of Phase 0, when legacy claims data profiling surfaced fragmentation materially worse than the estimate the BRD-01 schedule was built on.

TIER 2 · UNDER REVIEW Submitted 10 September 2026, pending Executive Steering Board and AI Governance Board decision at the 25 September session. No baseline has changed — the cost, schedule, and scope baselines remain as authorized until this request is approved or rejected.

1 · Request Header

Change Request IDCR-003
TitleLegacy claims data remediation — additional scope and effort to support BRD-01 model training
Change typeScope addition with cost and schedule impact
TierTier 2 — exceeds Program Director standing authority (scope change of any size; budget > $500K)
StatusUnder review
Date raised10 September 2026 (Week 4, Phase 0)
Requested byData & Cloud AI Platform Lead
Change ownerC. Tyrrell, Program Director
Approval authorityExecutive Steering Board (cost/schedule) with AI Governance Board endorsement (model data quality)
Target decision date25 September 2026, alongside the AI Readiness Assessment Report
Linked riskRSK-01 — Legacy claims data fragmentation (inherent rating: Likely × High)

2 · Background & Description of Change

The Phase 0 AI Readiness Assessment includes a data-quality profiling exercise across the claims and prior-authorization source systems that will feed BRD-01. Profiling is substantially complete and has established that source data is materially more fragmented than the planning assumption used to build the BRD-01 schedule.

Specifically, profiling has found: inconsistent provider identifiers across three claims platforms arising from historical acquisitions; incomplete diagnosis and procedure coding on a significant proportion of older records; no reliable linkage key between prior-authorization requests and their resulting claims; and inconsistent handling of adjustment and reversal records. Individually each is manageable. Together, they mean that the training dataset assembled from these sources would not be fit to train a prior-authorization decisioning model to the standard the program's Master Test & QA Strategy requires — specifically dimensions 1 (data quality and representativeness) and 5 (fairness), since coding gaps are not evenly distributed across the member population.

This request seeks approval for a dedicated data remediation workstream — additional engineering effort, tooling, and elapsed time — to establish a governed, linked, quality-assured claims dataset before BRD-01 model development begins, rather than discovering the problem during model training.

3 · Reason for Change & Consequence of Not Acting

The program's planning assumption was that source claims data would require standard cleansing within the effort already provided in WBS element 2.0. Profiling has disproved that assumption. This is not a scope preference; it is a correction to a factual estimate, surfaced by the Phase-0 activity designed to surface exactly this class of problem.

If the change is not approved: BRD-01 model development would proceed on a dataset known to be unfit. The most likely outcome is that the model fails independent validation on data quality or fairness grounds, forcing rework at a far later and more expensive point — after model development, integration, and testing effort has already been spent. Secondary consequences include a materially higher risk of a fairness finding on a member-facing prior-authorization decision, which carries regulatory exposure under the same scrutiny anticipated in RSK-04.

4 · Options Analysis

OptionDescriptionCostSchedule ImpactResidual Risk
Option A
Do nothing
Proceed on existing data with in-flight cleansing during model development$0None initially; high probability of 3–5 month rework laterUnacceptable — near-certain IMV failure on data quality and fairness
Option B
Partial remediation
Remediate identifier linkage only; accept coding gaps and handle statistically during modelling$740,0003 weeks to BRD-01 requirements sign-offModerate — linkage resolved, but coding gaps still create sub-population bias exposure
Option C
Full remediation Recommended
Dedicated remediation workstream: identifier resolution, coding completion where recoverable, PA-to-claim linkage, adjustment/reversal normalization, plus a permanent data-quality rules engine$1,240,0006 weeks to BRD-01 requirements sign-off; no impact to the 30 Sep 2027 production milestone (absorbed within BRD-01 float)Low — dataset fit for training and for fairness assessment
Option D
Defer BRD-01
Re-sequence so BRD-02 leads and BRD-01 follows after remediation$2,100,000Re-baselines the entire program; BRD-01 production slips ~9 monthsRejected — sacrifices the flagship, disrupts CMS-0057-F coordination, poor value

5 · Impact Assessment

Assessment below is for the recommended Option C.

DimensionImpactDetail
ScopeAdditionNew deliverable "Legacy claims data remediation" added under WBS 2.0 Data & Cloud AI Platform. No existing scope removed.
Cost+$1,240,000Funded from contingency reserve, not from a budget increase. Reserve reduces $9.0M → $7.76M (7.8% of authorized baseline, remaining within acceptable range for the risk profile).
Schedule6 weeksBRD-01 requirements sign-off moves 18 Dec 2026 → 29 Jan 2027. Platform Foundation (12 Mar 2027) and BRD-01 Production (30 Sep 2027) milestones are unchanged — the six weeks are absorbed within existing float on the BRD-01 path.
Resources+6 FTE, 14 weeks4 data engineers (offshore, synthetic/de-identified work), 2 onshore data analysts for PHI-bearing linkage work. Sourced through existing Pulaski SOW-01 capacity; no new vendor.
QualityPositiveDirectly improves the evidence base for test dimensions 1 and 5. Reduces probability of an IMV failure at the BRD-01 production gate.
RiskReduces RSK-01Moves RSK-01 from Likely × High toward Possible × Moderate once remediation completes. Introduces minor new delivery risk around remediation duration, tracked as a sub-risk.
BenefitsNeutral to positiveNo change to the $27.2M steady-state benefit case. Protects the largest benefit stream (PA automation, $12.8M/yr) from a delayed or degraded launch.
CompliancePositiveA governed, lineage-traceable dataset strengthens the program's position under NAIC and state AI-in-insurance scrutiny.
ContractualNoneWithin SOW-01 scope and envelope; no amendment required. Milestone payment schedule unchanged.

6 · Recommendation

Recommend approval of Option C — full remediation, $1,240,000 drawn from contingency reserve. It is the only option that produces a dataset fit for both model training and defensible fairness assessment, and it does so without moving the program's external commitments: BRD-01 still reaches production on 30 September 2027. Option A defers a known problem into a more expensive phase; Option B leaves the fairness exposure — the one risk the program is least able to absorb — substantially open; Option D destroys the Year 1 value case for a problem that has a $1.24M solution. The contingency reserve was sized at 9.1% precisely because RSK-01 was identified as a leading exposure at authorization; this is the reserve performing its intended function.

7 · Implementation Plan (if approved)

StepActivityOwnerDurationCompletion
1Mobilize remediation team; confirm onshore/offshore data-access boundaries with Data Privacy OfficePlatform Lead / HR1 week2 Oct 2026
2Provider identifier resolution across the three claims platformsData engineering4 weeks30 Oct 2026
3Prior-auth to claim linkage key construction and validationData engineering3 weeks20 Nov 2026
4Diagnosis/procedure coding completion where recoverable; gap documentation where notData analysts + Clinical4 weeks18 Dec 2026
5Adjustment and reversal normalizationData engineering2 weeks8 Jan 2027
6Data-quality rules engine deployed; sub-population coverage verified for fairness testingPlatform + IMV2 weeks22 Jan 2027
7Dataset governance sign-off; BRD-01 requirements sign-offAI Governance Board1 week29 Jan 2027

8 · Affected Baselines & Artifacts

If approved, the following are updated under configuration control within five business days of the decision:

ArtifactChange required
Program BudgetContingency ledger entry: $1.24M drawn, reserve balance $7.76M. Category and workstream views updated; total authorized baseline unchanged at $99.0M.
WBS Budget Rollup & Resource-Loaded WBSNew deliverable and seven work packages under element 2.0; resource assignments and dates added.
Program Management PlanMilestone table: BRD-01 requirements sign-off 18 Dec 2026 → 29 Jan 2027.
RAIDD LogRSK-01 updated to "mitigation in execution" with revised residual rating; new sub-risk for remediation duration.
BRD-01Data requirements section updated to reference the remediated dataset and its governance sign-off.
Change Control LogCR-003 status moved to Approved with decision date and approver record.
Program DashboardContingency drawdown reflected; RSK-01 heat-map position updated at next reporting cycle.

9 · Approval Routing & Decision Record

Tier 2 changes require the relevant board's approval. Because this request affects both the cost baseline and model data quality, it is routed to both the Executive Steering Board and the AI Governance Board, with Program Finance and SOX review preceding the decision.

Reviewer / ApproverRole in decisionStatusDate
Program Finance (A. Rodriguez)Cost validation and contingency impactReviewed — supports12 Sep 2026
SOX / Financial ControlsControls review (contingency drawdown)Reviewed — no objection15 Sep 2026
Data Privacy Office (E. Sato)PHI access boundaries for remediation workReviewed — conditions noted16 Sep 2026
Chief Enterprise Architect (D. Chen)Architecture impact — data-quality rules engineReviewed — supports17 Sep 2026
AI Governance Board (S. Khurana)Endorsement — model data quality and fairness basisScheduled25 Sep 2026
Executive Steering Board (Chair: M. Kavanagh)Approval authority — cost and schedule baselineScheduled25 Sep 2026

Decision block

Decision☐ Approved    ☐ Approved with conditions    ☐ Rejected    ☐ Deferred
Conditions (if any) 
Executive Sponsor signature 
Date 

10 · Audit Trail

DateEventActor
08 Sep 2026Profiling findings escalated from the AI Readiness Assessment workstreamData & Cloud AI Platform Lead
10 Sep 2026CR-003 raised and logged; tier assessed as Tier 2PMO (T. Valdez)
11 Sep 2026Options analysis and impact assessment draftedProgram Director + Platform Lead
12–17 Sep 2026Functional reviews completed (Finance, SOX, Privacy, Architecture)Respective functions
25 Sep 2026Scheduled for joint ESB / AIGB decisionPMO
How this fits the process. Tier 1 changes are decided by the Program Director within standing authority and recorded on the log without a document at this depth. Tier 2 — this request — requires the full analysis above and a board decision. Tier 3 (contingency draws above $1M in aggregate, program viability, or Charter re-baseline) escalates to the Executive Sponsor and CFO. The tier definitions and turnaround targets are set out in the Change Control Log.