Anonymized engagement summary: generative AI for contract audits and policy drafting at a public transit district (2023)
Contracted — countersigned (EPC-reported; documents private)First-pass editionA public transit district (client not named) · 2023 · Project ID EPC-2023-ANON-TRANSIT-GENAI
Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Opening answer. In May 2023 a public transit district contracted EPC Group for a 40-hour engagement to plan and implement a generative AI capability for auditing its contracts and drafting content such as policies and procedures, with up to five Power Automate flows, beta tests on the district's own contracts and go-live support. The statement of work was countersigned by both parties on May 23, 2023.
Disclosure. This is an EPC-reported anonymized engagement summary. The client is not named. Supporting project records — the countersigned statement of work — are retained privately and are not published in the Evidence Center, so this account is outside the Evidence Center's documented-engagement counts. Every engagement fact below is taken from that statement of work; nothing is added from memory.
Verification class
Contracted — countersigned (EPC-reported; documents private). The statement of work carries both signatures dated May 23, 2023: EPC Group's founder and the district's general manager. That establishes the contracted scope. It does not establish delivery, acceptance or any result, and no such document is cited here.
Buyer problem and scope
| Item | What the statement of work states |
|---|---|
| Buyer | A public transit district (client not named) |
| Objective | Implement what the statement of work calls "Azure CoPilot" to assist with contract audits and with creating content such as policies and procedures |
| Corpus | About 2,000 pages, as stated in the desired outcomes |
| Desired outcomes | An incoming-parameters list; outgoing requirements for the AI output; up to five Power Automate flows; beta tests on the district's contracts; go-live support |
| Engagement size | A 40-hour block of time; the statement of work notes the client understood that AI work could require more than 40 hours for final deliverables on the first building block |
| Phases | Phase 1, discovery and assessment: business goals, compliance and security requirements, and remedial or pre-deployment tasks. Phase 2, design: how each component would be implemented against the requirements |
| Deliverables | Weekly project schedule reports (status, activities, problems, recommendations) and technical recommendations: configure and implement the AI capability according to the district's technical requirements |
| Roles | A project manager (single point of contact, status reporting, schedule tracking) and Microsoft solutions architects (assessment, requirements alignment, provisioning and configuration, integrations, quality assurance and testing, post-deployment support) |
| Acceptance terms | Milestone service deliverables submitted for the client's review, with an acceptance-form process — a contract term, not evidence that acceptance occurred |
| Dates | Statement of work effective May 23, 2023; both signatures dated May 23, 2023 |
Unknowns, not documented in the record reviewed: which Azure AI service and model were finally configured, the number of contracts processed in beta testing, whether all five flows were built, the go-live date, hours consumed against the 40-hour block, and any measured effect on audit time or policy drafting.
Approach and artifacts
The statement of work describes a two-phase plan. Discovery and assessment establishes the district's goals, its compliance and security requirements and any remediation needed before deployment; design then determines how each component is implemented. The desired outcomes add up to five Power Automate flows and beta tests on the district's own contracts before go-live support; the statement of work does not specify what each flow does.
No artifact from execution — a parameters list, a flow export, a test record or a go-live log — is cited here. EPC Group's AI consulting services page describes how assessment-to-production AI engagements are run today.
Results and limitations
| Claim | Basis | Status |
|---|---|---|
| The district contracted a 40-hour generative AI engagement for contract audits and policy drafting | Countersigned statement of work, May 23, 2023 | Contracted (EPC-reported; document private) |
| The scope included up to five Power Automate flows and beta tests on the district's contracts | Statement of work, desired outcomes | Contracted scope |
| The capability was implemented and went live | No document cited | Not evidenced |
| Audit or drafting time changed | No measurement cited | Not evidenced |
The product name "Azure CoPilot" is quoted as the 2023 statement of work wrote it. It is not a current Microsoft product name, and this summary does not claim which Azure AI service was finally used.
Provenance
- Source: the countersigned statement of work (10 pages, both signatures dated May 23, 2023), held privately by EPC Group. Engagement ID: EPC-2023-ANON-TRANSIT-GENAI, an alias unrelated to the client's name.
- Anonymization: industry and organization type only, under EPC Group's anonymized-case rule; commercial terms are not published.
- Not claimed: delivery, acceptance, client endorsement or any measured outcome.
- Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group.
Revision history
- 2026-10-25: anonymized engagement summary published from the countersigned statement of work.
Related service
AI consulting services · Talk to a senior architect
Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group · Published: 2026-10-25.