New features let districts, schools and teachers build and customize AI tools around their policies, curriculum and instructional needs.
SEATTLE, WA, UNITED STATES, October 23, 2026 /EINPresswire.com/ — For years, enterprise teams have enjoyed the freedom to construct bespoke AI tools on top of commercial platforms. K-12 schools, however, have been stuck choosing from a limited menu of pre-packaged options. That dynamic shifted this week, as Colleague AI, the research-backed K-12 platform supported by $19M+ in federal grants from the NSF, DOE, and IES, unveiled what it calls a first: the ability for districts, schools, and teachers to build, clone, and customize the AI Assistants that power the platform, rather than settling for standard issue offerings.
Previously, Colleague AI’s “one platform, every stakeholder” model meant a fixed suite of AI Assistants designed for teachers, students, schools, and districts. Now, the framework behind those tools is transparent and editable at every level of the organization. A district can examine the AI Assistants in its instance, clone an existing one, or create a new tool from scratch, tailoring it to a specific policy, curriculum, or student population that no generic template adequately addresses.
“This is the moment K-12 catches up to where other industries already are — without asking a single teacher to become an engineer to get there,” said Dr. Min Sun, co-founder and CEO of Colleague AI. “For the first time, a district, a school and a teacher can build the AI tools they actually need, inside a platform built for education from day one, without ever leaving the research and governance foundation their board already trusts.”
“This is agentic architecture built for a classroom, not bolted onto one,” said Jian (Kevin) He, co-founder and CTO. “A district, school or teacher can give their AI Assistant its own skills and tools and benefit from the fine-tuning and RAG system that holds district documents, policies, curricula, and information. Those AI Assistants can have sub-agents of their own, making Colleague AI the first K-12 platform built to let schools design and deploy real multi-agent systems — and every one of them inherits the same FERPA, COPPA, GDPR and SOC 2 posture already documented in our data processing agreements.”
Beyond customization, the platform introduces a knowledge graph designed to give it persistent memory of the teachers and students it serves. This memory tracks learner progression, teacher professional growth, and organizational improvement. The system uses this context to shape outputs — an assignment, lesson, or feedback piece informed by a student’s specific strengths, gaps, accommodations, and pace — rather than relying on generic, non-personalized content. The result is that AI output improves over time as it learns, rather than starting from scratch with each request, a limitation of current competitors. This approach aims to enable personalized learning at scale, providing learning opportunities for all students without adding to teacher or administrative workloads.
For school and district clients, a new administrative dashboard offers a consolidated view of every AI Assistant and sub-agent in use, including who built it and its usage patterns. Early adopters are already constructing a school improvement plan assistant from their own data, a grading-consistency checker calibrated to their own rubric language, and an IEP drafting tool based on their own documentation — capabilities that no fixed-feature-set or general-purpose AI account could offer.
For teachers, this means tasks that previously required toggling between a lesson planner, a grading tool, and a newsletter builder can now be handled in a single conversation, with each task informed by the context of a specific student or class. Colleague AI frames this as a move toward 10× output, cutting down repetitive setup while keeping teachers in charge of instructional judgment. One educator shared that adapting an entire unit into project-based learning took roughly 100 hours before using AI, then around 12 hours on average with AI. With Colleague AI’s new features, this specific workflow took 30 minutes. “I can spend my time on the stuff that actually needs me: Is this accurate? Is it rigorous? Will my kids care?”
For district leaders, the same approach supports building and maintaining tools in-house, including school improvement plans and intervention workflows tailored to local needs, rather than waiting for a vendor’s roadmap or an outside consultant.
The new product features will be released to enterprise clients later this month before being rolled out across the user base to paid and freemium users. Districts and schools interested in the Colleague AI platform with its governed and customized AI tools and Assistants can learn more at colleague.ai.
About Colleague AI
Colleague AI is the leading comprehensive, research-backed K-12 AI infrastructure platform serving teachers, students, and school leaders. The platform services the entire teaching and learning cycle, offering unlimited educator and administrative tools, easy-to-use embedded workflows, and AI management infrastructure. Built with more than $19 million in federal research funding from the NSF, U.S. Department of Education and Institute of Education Sciences, and developed in partnership with AmplifyLearn.AI at the University of Washington College of Education, Colleague AI serves 300K+ users nationwide and in 50+ countries. For more information, visit colleague.ai.
Naomi Thomas
Colleague AI
naomi.thomas@colleague.ai
Visit us on social media:
LinkedIn
Facebook
YouTube
X
—
**Why it matters:** This move signals a shift in K-12 technology from a “one-size-fits-all” model to one where educators hold the reins of their AI tools. By empowering districts to build within a governance framework, it could accelerate the adoption of personalized learning while addressing concerns about data privacy and policy compliance.
Also worth a look: Cobalt AI report: Security teams drowning in alarms while budgets shrink, 71% cite overload and Vitek IP puts CodeScribe’s medical simulation patent portfolio on the market.
