From the Microsoft edition of November 1, 2026
Chelan, Wash., drafts an AI use policy that makes Microsoft Copilot the preferred tool, reserves an enterprise Claude for complex work, and treats AI prompts and outputs as potential public records (council first look Oct. 27; reported Oct. 31)
City of Chelan · Local government · Chelan, Washington, United States
- Publisher
- Lake Chelan Mirror (Ward Media)
- Original publication
- 2026-10-31
- Source retrieved
- 2026-11-02
- Event date
- 2026-10-27
What happened
The Lake Chelan Mirror reports that Chelan's council got its first look at a draft AI use policy on Oct. 27 and took no action. Microsoft Copilot would be the preferred platform because it works inside Outlook, Word and Excel. Free Copilot could be used only with publicly available information; licensed enterprise Copilot could be used with authorised city information. For complex tasks such as technical analysis, the city would favour an enterprise version of Claude. Personal ChatGPT and Gemini accounts would generally be barred. AI could not be the sole basis for permitting, enforcement or employment decisions. Prompts, conversations and outputs may be public records, and qualifying records must be kept in approved city systems.
Why it matters
A small US city is writing a tiered Microsoft policy: free Copilot for public data, enterprise Copilot for city data, and a second vendor for heavier work. It also treats Copilot conversations as records under Washington public-records law. That is the decision many small local governments face when Copilot Chat arrives inside Microsoft 365.
Evidence and measured results
A local reporter's account of a staff presentation and council discussion. The draft policy text was not published with the article and was not read.
Limitations and uncertainty
Draft only; council took no action and the policy may change. Based on news reporting, not the policy text. A small city, so scale and staffing differ from large jurisdictions. The Claude element is outside this stream and is noted only for context.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-02; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- A small city wants staff to use AI that is already in Microsoft 365 but must protect sensitive data and meet public-records law.
- Stakeholders
- City administrator
- HR and communications director
- IT staff
- City attorney
- City council
- Discovery questions
- Which tasks does the city expect Copilot to handle, and which go to another tool?
- How will staff tell tenant-protected Copilot from consumer Copilot?
- How will AI prompts and outputs be found for a public-records request?
- Who approves new AI tools?
- Value hypothesis
- Helping the city stand up a tiered Copilot policy with records procedures may reduce unapproved consumer AI use, if staff adopt the approved path.
- Potential engagement
- A short policy-to-practice engagement: tenant settings review, training module and records-capture procedure for Copilot.
- Unsupported claims
- That Copilot replaces the separate tool the draft names for complex work
- That the policy has been adopted
- Any productivity gain for city staff
Pre-sales engineering
Role takeaway
- Fit
- Good fit for a Microsoft 365 city that wants governed Copilot Chat with clear data tiers.
- Architecture and integration
- Copilot Chat in the city's tenant, with Purview retention and eDiscovery for AI interactions; a second AI vendor would sit alongside with separate controls.
- Prerequisites
- City accounts for all staff using AI
- Sensitivity labels for confidential and law-enforcement content
- Retention policy covering Copilot interactions
- Constraints
- Small IT staff
- Washington public-records obligations
- Policy is still a draft
- Security
- Block or separate consumer Copilot, label attorney-client and law-enforcement content, and log AI use for audit and records requests.
- Proof of value
- Run a mock public-records request for Copilot prompts and outputs
- Check that labelled confidential content is excluded from Copilot answers
Delivery
Role takeaway
- Work
- Turn the draft policy into tenant settings, training and a records-capture procedure for Copilot.
- Dependencies
- Council adoption of the policy
- IT security review
- City attorney guidance on sensitive information
- Ownership
- City administrator for policy; IT for tenant controls and accounts.
- Skills and adoption
- Short role-based training on approved tools, verification and records duties before access is granted.
- Governance checkpoints
- Council adopts the policy
- Tenant settings match the policy's data tiers
- Annual policy review
- Acceptance criteria
- Proposed: 100% of AI users complete training before access
- Proposed: a test records request returns Copilot content within the statutory response process
- Proposed: no consumer AI accounts used for city business in a sampled audit
- Risks
- Policy changes before adoption
- Staff keep using consumer tools
- AI records lost before capture
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?
The free/enterprise split depends on users signing in with city accounts and on IT being able to tell tenant-protected Copilot apart from consumer Copilot. Dual-vendor use (Copilot plus Claude) means two sets of identity, logging and data controls.
Governance
Who approves, reviews and stays accountable for outcomes?
The city administrator would approve tools and IT would run security reviews. Annual policy review, mandatory training and graduated discipline are proposed. Human review is required before AI output is released or relied on.
Security and privacy
What data, permissions and controls need testing?
SSNs, financial and medical data would be limited to approved enterprise systems. Attorney-client, executive-session and some law-enforcement information would need the city attorney's authorisation before entry into any AI system.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Training on approved tools, security and public-records duties would be required before use.
Procurement
What should contracts, pricing and exit terms secure?
Copilot is preferred because it is bundled with existing Microsoft software, while a separate enterprise Claude purchase is contemplated for complex work, so Microsoft is not assumed to be the only AI vendor.
Operating model
Which teams own the service once it runs?
IT owns security review and accounts; the city administrator owns tool approval; employees own verification and records capture.
Publication history
- 2026-11-01Microsoft · Issue 043 resources
Stable resource ID: chelan-wa-draft-ai-policy-copilot-preferred-claude-for-complex-ai-prompts-public-records-20261031