This is a directional technical sequence with explicit gates, risks, and owners. It is not a fixed-price Phase 2 proposal and contains no implied staffing or delivery commitment.
Dates follow an approved baseline, scope, and staffing commitment. Sequence and exit gates are useful now; commercial commitments are not inferred.
One accepted Assessment Report with a measured baseline—not a collection of preliminary artifacts.
Five customer journeys work consistently through one supported MCP surface.
The proven demonstration can be operated safely for real partners and real traffic.
This page does not commit a fixed total, price by milestone, staffing plan, delivery date, or conference delivery. Those fields require approved scope, baseline, risk allowance, and authorized commercial owners.
This shows where work creates confidence—not a work breakdown, price, or committed schedule.
This is a feasibility statement, not a committed Phase 2 delivery date.
A conference-ready demonstration is technically feasible if scope is frozen around proven journeys and the evaluation gate closes in time.
If live access becomes unreliable, use stored redacted evidence from the same frozen release and state that the replay is evidence-backed rather than live.
Fallback evidence must come from the same frozen release and remain clearly labeled.The SOW excludes production hardening from Phase 1, but the assessment must identify the gaps, target state, owner, and delivery phase.
| Control area | Current state | Target state | Owner |
|---|---|---|---|
| SecurityBefore further external use | Read-only service controls exist; the temporary validation credential still requires rotation and final secret review. | Rotated secrets, repository/artifact scan, least privilege, and an approved test/operating policy. | Tenexity + OptiCat security |
| AuthenticationProduction hardening | The private MCP path uses an internal bearer boundary; production partner identity is not designed here. | Supported partner authentication with rotation, revocation, and documented trust boundaries. | OptiCat platform owner |
| AuthorizationProduction hardening | Catalog visibility depends on upstream key entitlement; customer/tenant isolation is outside the demonstration scope. | Per-partner authorization, entitlement visibility, quotas, and isolation tests. | OptiCat platform + API owner |
| ObservabilityDemo gate → production | Runs retain tool evidence, statuses, counts, and selected latency; production monitoring is not complete. | Health, latency, error, entitlement, truncation, and release metrics with operational ownership. | Tenexity engineering then OptiCat operations |
| LoggingProduction hardening | Redacted evaluation traces exist, but retention, access, and audit policy are not finalized. | Structured redacted logs, correlation IDs, retention controls, and incident-safe access. | OptiCat operations/security |
| AI governancePhase 1 closeout and every release | Evidence-only rules and KPI definitions exist; the full domain-approved replay remains open. | Versioned cases, named adjudicators, zero-tolerance safety gates, drift review, and controlled releases. | OptiCat product/domain + Tenexity evaluation |
| ID | Risk and impact | Response | Owner |
|---|---|---|---|
| R-01 | Key scope makes valid catalog data appear absent.False negatives and incorrect root-cause conclusions. | Record written entitlement, key-scope ID, health result, and complete paging with every run. | OptiCat API owner |
| R-02 | Historical expected answers have drifted or were never authoritative.A correct current result can be graded as a regression. | Require named domain adjudication and retain a visible data-drift outcome. | OptiCat catalog owner |
| R-03 | A polished demo is mistaken for a measured baseline.Investment or release decisions rely on representative success instead of repeatable evidence. | Keep validation, baseline, and production-readiness statuses separate everywhere. | Tenexity evaluation |
| R-04 | October scope expands faster than evidence can be approved.Conference reliability falls while feature count rises. | Freeze the required journeys, rehearse twice, and maintain evidence-backed fallback examples. | OptiCat product owner |
| R-05 | Commercial terms are inferred from the technical roadmap.An estimate is treated as a fixed commitment without approved staffing or contingency. | Keep price, milestone allocation, staffing commitment, and delivery date explicitly pending. | Authorized commercial owners |
| ID | Decision and rationale | State | Owner · reversibility |
|---|---|---|---|
| D-01 | Retain the OptiCat catalog connection and task-level MCP direction.Live catalog journeys prove the core is viable; the principal gaps are evidence fidelity and evaluation. | Decided | Assessment recommendationYes, but no replacement case is established |
| D-02 | Treat the site and demo as validation evidence, not the Phase 1 software deliverable.The SOW names one written assessment and explicitly excludes a software deliverable. | Decided | TenexityOnly by written scope change |
| D-03 | Publish no current accuracy percentage until the frozen replay is adjudicated.Historical labels came from different moments and cannot support one current score. | Decided | Tenexity evaluationNo; evidence standard |
| D-04 | Approve fitment-critical qualifiers and relationship semantics by category.The service can preserve data, but domain experts must decide what changes a safe answer. | Pending | OptiCat domain ownersVersioned as domain policy evolves |
| D-05 | Freeze the October demonstration journeys and fallback behavior.Conference readiness needs a bounded story and a named acceptance owner. | Pending | OptiCat product/executiveYes, before scope freeze |
| D-06 | Authorize a fixed-price Phase 2 proposal.Commercial terms require an approved baseline, scope, staffing plan, risk allowance, and delivery commitment. | Pending | Authorized commercial ownersYes, until contract execution |