evidoria

← Back to browse

Good practice Imported

Colleague AI — University of Washington's Multi-District Generative AI Classroom Pilot Scales to 4,600+ Teachers Across Puget Sound

United States of America · Seattle · See the United States of America profile · See the Seattle profile

Evidence: Descriptive / self-reported Top 56% 47/100 · Ask Evidence Copilot about this practice

A University of Washington research center's Colleague AI classroom platform grew from a 21-teacher, seven-week pilot in five Puget Sound schools to over 4,600 teachers across 12 Washington districts by March 2026, used mainly for AI-assisted lesson planning.

21 teachers
Teachers in the spring 2025 pilot (spring 2025)
600+ students
Students in the spring 2025 pilot (spring 2025)
1,438 teachers
Teachers using the platform, September 2025 (September 2025)
2,891 teachers
Teachers using the platform, December 2025 (December 2025)
4,644 teachers
Teachers using the platform, March 2026 (March 2026)
412,304 messages
Messages exchanged with the AI assistant (through March 2026)
60–99 %
Share of platform usage that is lesson planning (varies by district)
Colleague AI — University of Washington's Multi-District Generative AI Classroom Pilot Scales to 4,600+ Teachers Across Puget Sound

Details

Maturity
Scaling
Promoter
AmplifyLearn.AI, University of Washington College of Education
Period
Spring 2025 pilot, scaled through March 2026
Keywords
K-12 education, generative AI, teacher professional development, edtech research

Context

AmplifyLearn.AI, a University of Washington College of Education research center backed by a $10 million grant to develop generative-AI teaching tools, partnered with the startup Colleague AI to build a classroom platform used directly by teachers and students.

Objectives

Test and scale a generative-AI classroom platform (Teaching Aide, Assessment and AI Grading, AI Tutor, and Student Growth Insights) for lesson planning, grading, tutoring and growth insights across K-12 classrooms.

Activities

In spring 2025, 21 in-service teachers across four Washington public school districts and one independent school piloted the platform for seven weeks with more than 600 students in grades 6-12. By the 2025-26 school year, a mid-year report covering 12 Washington districts recorded 1,438 teachers using Colleague AI in September 2025, growing to 2,891 by December 2025 and 4,644 by March 2026 (a 61% rise since December), exchanging over 412,304 messages with the AI assistant; lesson planning accounted for 60-99% of platform usage in every participating district.

Results

Researchers link platform usage to district-held demographic, attendance and test-score records to study adoption patterns by teacher experience, credential status and classroom composition.

Conclusions

The report is an explicitly descriptive mid-year account of adoption, not a causal study of learning outcomes, and no student achievement results have been published yet.

Implementation
Indicative cost
High (€500k–€5M) — Underpinned by a $10 million grant to AmplifyLearn.AI; scaled across 12 Washington districts of varying size and setting (rural, suburban, urban) within about a year.
Time to results
Medium (1–3 years) — Seven-week pilot in spring 2025 with 21 teachers, scaling to 4,644 teachers across 12 districts by March 2026 (a 61% rise since December 2025).
Staffing & skills
AmplifyLearn.AI, University of Washington College of Education, Colleague AI (platform partner/startup)

Conditions for success

  • Backed by a $10 million research grant enabling sustained platform development and district partnerships
  • Linking usage data to district-held demographic, attendance and test-score records to study adoption patterns by teacher experience and credential status

Common failure modes

  • The report is explicitly descriptive of adoption, not a causal study; no student-achievement results have been published yet
  • Usage is concentrated heavily in lesson planning (60-99%) rather than the platform's other advertised features

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Replication kit

Reusable artefacts from this practice — as published by their sources.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful