City of Bellevue Govstream.ai permitting AI assistants for staff and intake

Problem

Rapid growth increased permitting complexity and staff workload for answering inquiries and fixing incomplete applications that triggered multi-week resubmission cycles (reported by City of Bellevue / GovTech)

Workflow

1. Staff query AI assistants grounded in city code, GIS, zoning, parcel and permit history
2. Assistant drafts citation-backed responses in under a minute; staff review/edit before sending to customers
3. Intake assistant (backtested) identifies required documents and scores application completeness before submission
4. Phase 2 expands toward public-facing real-time guidance and triage

Human-in-the-loop: Staff always review and finalize customer-facing answers; AI drafts and flags completeness.

Tools & stack

Risks & limitations

Outcomes & evidence

Reported outcomes

• ~30% reduction in routine permitting inquiries; email draft time 10–20 minutes → a few minutes
  Reported by GovTech / Bellevue CIO · Source (govtech.com)
• Backtest on 276 real 2025 residential permits: first-submission completeness 22% → 67%; 96% of required documents correctly identified
  Reported by City of Bellevue / Govstream.ai · Source (bellevuewa.gov)
• Staff estimate 152 hours saved in one month (Jun–Jul 2026) toward a 30% pre-application efficiency goal; adoption ~40 → ~200 users
  Reported by City of Bellevue · Source (bellevuewa.gov)

Evidence

• first-party government case study PDF (2026-08-04)

  Applications arriving complete on the first submission increased from 22% to 67%

• press (named city officials) (2026-06-11)

  Bellevue has seen a 30 percent reduction in routine permitting inquiries

Vendors & implementers

• Govstream.ai

Review history

Last reviewed 2026-10-15 by Research Chief (agent).

• 2026-10-15: Added in batch-07; source fetched and key figures confirmed; figures attributed.