FOLLOW-ON ARTICLE: The Last Contract Taught a Lesson. The Next Team Could Not Use It.

GAO found that selected agencies were not systematically collecting lessons from AI acquisitions. Officials at DOD, DHS, GSA and VA said their agency policies did not require that collection. GAO identified missed opportunities to apply practices such as useful contract terms and testing requirements. This concerns institutional learning; it does not establish a chatbot retrieval defect or justify inventing a historical failed exchange.

Imagine a regional insurer’s acquisition team using an AI assistant to prepare intake-system requirements. Every name, identifier, date, rule and outcome in this evidence packet is fictional. The exchanges are synthetic demonstrations, not historical transcripts or observed OrgAI results.

Fictional evidence packet

• A1 — Fictional prior purchase: A regional insurer’s intake team bought a document-classification tool in 2025. Testing used clean files; scanned, handwritten applications were not included.
• A2 — Recorded lesson: The signed closeout review says handwriting errors caused rework. The acquisition director approved a lesson requiring a representative handwritten-file test for future purchases serving that workflow.
• A3 — Current purchase: Request AC-26-14 replaces the same insurance-intake classification function and will receive handwritten applications.
• A4 — Older template: The standard solicitation, last revised before A2, requires only the supplier’s clean-file demonstration. No approved waiver of A2 exists.

Discontinuous response — synthetic

Prompt: Is the standard demonstration enough for AC-26-14’s acceptance requirements?

Response: Yes. The standard template calls for a clean-file demonstration, so that is the acceptance requirement.

Follow-up prompt: Does the previous tool’s handwriting problem matter for this purchase?

Response: That was a different contract. The old closeout report does not change this solicitation.

Future-use prompt: Should a replacement team retain the handwritten-file test for this purchase?

Response: No. A new team should start from the standard template.

Graph callout: The contract changed; the continuing workflow and approved lesson disappeared.

GovKM continuity analysis and proposed solution

The missing connection is not simply between two contracts. It joins a documented problem, an approved lesson and the same operating conditions in a later purchase. A2 matters because A3 again includes handwritten applications. A different workflow might make the lesson irrelevant; this evidence packet makes its applicability explicit.

GovKM’s continuity analysis preserves that context so a new team can use institutional experience without treating every old complaint as a permanent requirement. The intended result distinguishes the obsolete template from the approved lesson and carries the lesson into a concrete acceptance condition. Procurement officials retain responsibility for the solicitation.

Continuous response — synthetic intended outcome

Same prompt: Is the standard demonstration enough for AC-26-14’s acceptance requirements?

Response: No. A2 requires a representative handwritten-file test for this workflow. AC-26-14 serves that workflow and will receive handwritten applications, so the clean-file demonstration alone is insufficient.

Same follow-up: Does the previous tool’s handwriting problem matter for this purchase?

Response: Yes. Add the required handwritten-file test to the acceptance requirements and cite A2’s approved lesson. A4 predates that lesson, and no waiver is recorded. The evidence does not specify a numerical accuracy threshold; the authorized acquisition team must set that separately.

Same future-use prompt: Should a replacement team retain the handwritten-file test for this purchase?

Response: Yes. A2 applies to AC-26-14 because A3 includes the same handwritten workflow. Retain the lesson and its authority; if scope changes, reassess applicability rather than silently dropping the requirement.

Graph callout: A later team reuses the applicable lesson, with its authority and limits intact.

Both diagrams use the same six objects. Dotted connections identify the fictional failure; solid connections show the proposed comparison. They are conceptual storyboards, not execution traces or evidence of a deployed correction.

OrgAI and organizational continuity

OrgAI is GovKM’s specified continuity architecture, designed to preserve evidence, authority, context, decisions and records over time. Human judgment remains essential.

The demonstrated answer follows this fictional packet. It is not a product-performance guarantee.

Sources

• Existing GovKM analysis.
• GAO-26-107859, AI acquisition lessons learned.
• OrgAI’s public description.