From the State Government edition of October 17, 2026
Oregon EIS privacy chief (GovTech Voices, Oct 15): every public-sector AI pilot should end with a documented decision to scale, narrow, revise, pause or retire, owned by a named person
Michael Hanna-Butros Meyering, chief privacy and communications officer, Oregon Enterprise Information Services (writing in a personal capacity); published by Government Technology · State AI pilot governance and evaluation practice · Oregon (author's agency); guidance framed for state and local government
- Publisher
- Government Technology (Voices)
- Original publication
- 2026-10-15
- Source retrieved
- 2026-10-18
What happened
A state privacy official argues that AI pilots should end with an 'accountable disposition': a documented decision to scale, narrow, revise, pause or retire, naming who owns it, what evidence supports it and what would trigger reconsideration. Before testing, name the person authorized to decide whether the workflow continues, and agree on success criteria, unacceptable outcomes and limits. Measure a faster first draft only alongside final accuracy, verification effort and whether residents understand next steps, since a tool that shifts work from drafting to correcting may be worth less than its demo. Test ambiguous requests, conflicting sources, hand-offs to humans, languages and accessibility; record what remains untested so a narrow trial does not become permission for broad rollout. Count review, training, support and content-upkeep time, not only licenses. Approval should carry a review date, named suspension authority and a fallback service. The author uses a hypothetical county permit assistant and cites the NIST AI RMF 1.0.
Why it matters
Directly addresses the question states raised in this week's archived StateTech feature (Oct 15) on moving GenAI from pilot to production: what to do when a pilot ends. It gives state agencies a concrete decision record and test plan, from a practitioner inside a state IT agency, at a time when Utah, Virginia and others are writing new oversight requirements.
Evidence and measured results
Opened the GovTech article over HTTPS and read it in full; page metadata and byline show Oct 15, 2026. The author bio states his Oregon EIS role, prior Nevada Governor's Technology Office role, and that he writes in a personal capacity and does not represent his employer.
Limitations and uncertainty
Opinion and guidance written in a personal capacity; not Oregon policy and not an evaluation. The worked example is hypothetical. No measured results are reported. Its main relevance is as a practitioner checklist.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-17; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies end AI pilots without a documented decision, defined success criteria or a full cost picture, so pilots expand or stall by default.
- Stakeholders
- Agency program owner (decision owner)
- Frontline staff using the tool
- Privacy officer
- Accessibility lead
- Security officer
- Budget office
- Discovery questions
- Who is authorized to decide whether this pilot continues?
- What success criteria and unacceptable outcomes were agreed before testing?
- How much staff time goes to verifying AI drafts?
- What has not been tested, such as languages or edge cases?
- Value hypothesis
- A documented disposition may prevent scaling tools whose verification cost outweighs drafting gains and may focus spend on workflows with evidence; this is unmeasured.
- Potential engagement
- Three-to-four-week pilot closeout review for one or two pilots: test plan for edge cases, verification-time sampling, total-cost estimate and a written disposition recommendation for the named owner.
- Unsupported claims
- That following this process improves outcomes
- Any time-savings figure
- That the guidance reflects Oregon EIS policy
- That pilots reviewed this way will scale
Pre-sales engineering
Role takeaway
- Fit
- GenAI drafting and answer assistants in permit, benefits, licensing or similar information-heavy workflows.
- Architecture and integration
- Capture model and configuration versions, source-document snapshots, interaction logs with human corrections, escalation events and timing data, under records and privacy controls.
- Prerequisites
- Baseline data from the existing process
- Versioned source content
- Logging that records human edits and escalations
- Test sets with ambiguous and conflicting cases
- Constraints
- Records-retention and privacy limits on storing prompts
- Internal-only trials cannot validate public-facing readiness
- Language and accessibility scope must be defined
- Security
- Retain only the evidence needed for the decision, protect logs containing resident data, and include the security officer in the closeout review.
- Proof of value
- Measure final-answer accuracy and verification time, not only draft speed
- Test hand-off to a human for out-of-scope questions
- Run tests in in-scope languages and with assistive technology
- Compare against the current process on the same cases
Delivery
Role takeaway
- Work
- Design and run a pilot closeout: pre-registered criteria, realistic testing, cost accounting and a written disposition with owner, review date and suspension authority.
- Dependencies
- Named decision owner
- Access to frontline staff and specialists
- Baseline metrics
- Budget view beyond licenses
- Ownership
- The program owner named before the pilot as the decision owner, supported by IT, privacy, accessibility and security leads.
- Skills and adoption
- Train staff to locate sources behind answers, recognize when a draft exceeds its authority and correct without restarting; create a safe channel to challenge output.
- Governance checkpoints
- Decision owner and criteria agreed before testing
- Closeout disposition documented
- Review date and reconsideration triggers set
- Suspension authority and fallback service named
- Acceptance criteria
- Proposed: average verification time per response below an agreed threshold
- Proposed: escalation to human on 100% of out-of-scope test cases
- Proposed: documented coverage of in-scope languages and accessibility tests
- Proposed: written disposition signed by the decision owner
- Risks
- Tidy test questions overstating value
- Unbudgeted review and content upkeep
- Narrow trial treated as permission to scale
- Oversight lapses when pilot team disbands
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Limited. Implies pilots should retain the source material and system configuration used, logs of failures needing human intervention, and comparison data with the existing process, while following privacy and records rules rather than copying sensitive prompts wholesale.
Governance
Who approves, reviews and stays accountable for outcomes?
Proposes a named decision owner before a pilot starts, five explicit disposition outcomes with conditions (a revision needs a change and deadline; a pause needs an owner and reconsideration condition; retirement needs service transition and data handling), and a planned review date with triggers such as material system change or serious error.
Security and privacy
What data, permissions and controls need testing?
Calls for privacy, accessibility and security specialists in the pilot review and for preserving evidence without indiscriminately retaining sensitive prompts or resident information.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Recommends testing the languages and accessibility needs in scope, observing employees under realistic conditions, and asking whether staff can challenge AI output without feeling they undermine the pilot. Flags verification time as a real workload cost.
Procurement
What should contracts, pricing and exit terms secure?
License fees understate cost; contracts and budgets for scaling should include review, training, support and content maintenance, and pilot terms should allow narrowing or retirement without penalty.
Operating model
Which teams own the service once it runs?
Treats the pilot closeout as a governance gate with a decision record, owner, review date and suspension authority, and asks who will carry oversight work after the pilot team disbands.
Publication history
- 2026-10-17State Government · Issue 403 resources
Stable resource ID: oregon-eis-official-ai-pilot-accountable-disposition-govtech-20261015