What the program is actually funding
Digital Promise's K-12 AI Infrastructure Program is seeking open-source tools that help teachers monitor and guide AI-enabled formative assessment, the ongoing, lower-stakes checking for understanding that happens throughout a lesson rather than at its end. The goal is to let teachers see what classroom AI tools such as tutors, grading assistants, content generators and analytics dashboards are actually doing, and to give them a direct way to redirect that behavior and adapt instruction in response.
The program will make up to two awards from a $1 million pool, with each grant capped at $600,000, and it explicitly excludes proprietary products that directly serve teachers and students from eligibility. Rebecca Griffiths, the program's Design Director, said the aim is to lower the barriers for edtech developers to build on shared infrastructure, increasing teachers' visibility into student interactions with AI and their ability to adapt for learner variability across a classroom.
The gap this is meant to close
The program names three specific failure modes it wants fixed: AI tools giving conflicting instruction to a student across different platforms, limited adaptation to different learners, and insufficient teacher oversight of what is actually happening inside an AI interaction in real time. These are not hypothetical problems. They are the exact complaints that have surfaced in teacher surveys over the past two years, where educators report feeling locked out of tools that are supposed to be helping them.
Funded projects must identify at least four concrete tasks that improve coordination between teachers and AI tools, explain how educators will test and refine the work throughout development, and show how the infrastructure will serve multilingual and neurodiverse students specifically rather than as an afterthought bolted on at the end. Applicants must also build a classroom demonstration application that actually engages teachers, not just a backend specification document that lives in a repository nobody outside the engineering team ever opens.
Who can apply, and who cannot
Most lead organizations may submit a single application, with separate provisions for divisions or departments inside organizations that have more than 2,000 staff, a structural detail that keeps the program from being dominated by one large institution submitting multiple competing proposals under different internal brands. Funding eligibility runs to US-based entities, including US territories, covering edtech companies, nonprofits, school districts and higher education institutions.
Nonprofits under fiscal sponsorship may also apply, as long as the resulting outputs clearly serve US-based students, and teams must pair AI or technical expertise with genuine education or learning sciences expertise rather than treating that pairing as optional. That requirement alone will filter out a fair number of purely technical teams that might otherwise have chased the grant money without a classroom-tested plan behind their proposal.
Why public infrastructure instead of another product
Most AI assessment investment to date has gone into proprietary products competing for district contracts, each building its own closed visibility layer that only works inside that one vendor's tool. Digital Promise is betting that the harder, less commercially attractive problem, shared infrastructure that any vendor's tool could plug into, is the piece the market will not fund on its own because no single company captures enough of the upside to justify building it alone.
That is a reasonable read of the market failure here. A teacher using three different AI tools across three subjects should not need three different dashboards to understand what each one is doing with a given student, and no single vendor has an incentive to build the shared standard that would make that possible. Public and philanthropic money filling that gap is a sound use of grant funding rather than a product subsidy in disguise.
The partners and the timeline
Digital Promise leads the program alongside Learning Data Insights, DrivenData, Georgetown University's Massive Data Institute and Catalyst at Penn's Graduate School of Education, a mix of research and data science partners that suggests the output will lean toward rigorous evaluation rather than a quick demo built to satisfy a grant deliverable. Applications close November 20, with recipients expected to be announced in early 2027, which gives applicants roughly a month from now to assemble a team that pairs AI or technical expertise with genuine education or learning sciences expertise, a pairing the program explicitly requires rather than treats as optional.
The grant period runs 18 months starting February 1, 2027, split into three phases: a six-month first phase building a working prototype from existing public goods, a second phase through summer 2027 that adds resources and evaluates improvements, and a third phase that refines the package and releases it publicly only after a safety review. That staged structure gives the field a checkpoint to course-correct before anything ships broadly, which is a more disciplined release process than most edtech procurement cycles ever see.
What district and L&D leaders should do now
If your district or vendor relationship already runs AI-enabled formative assessment, submit feedback or a letter of support during the application window, since the finished infrastructure will shape procurement conversations for years once it exists as an open standard other tools can adopt. Waiting until the 2027 release to engage means reacting to decisions made by applicants who took the time to apply and shape the specification now, while the requirements are still open for input rather than already locked into a funded proposal.
For corporate L&D teams watching the K-12 side for signal, this is the clearest evidence yet that teacher-facing AI oversight tools are becoming a funded category in their own right, distinct from the AI tools they are built to supervise. Expect comparable demand to surface in corporate training within 12 to 18 months, as compliance and learning teams start asking the same oversight question about AI coaching tools used with employees that Digital Promise is now asking about AI tutors used with students. The vocabulary will differ, dashboards instead of formative assessment, but the underlying governance problem is identical.
