Headline

AI in Education has moved past the “try it and see” phase in K–12, and districts are feeling that shift in real time. In a recent CoSN Podcast episode, host Stephen Gilfus brings together Dr. Tom Ryan and Narayan Nandigam (Nandi) to explore what’s changing, what leaders should prioritize, and what responsible adoption looks like when AI is already in teachers’ hands.
Listen Here

The CoSN PodcastAI in Education: From Experimentation to ImpactDecember 18, 2025 · 54 min 43 sPlay episode: AI in Education: From Experimentation to Impact (54 minutes 43 seconds)0:00-54:43Follow on SpotifyApple

From the opening minutes, Gilfus makes it clear this isn’t a conversation about a single product or pilot. Instead, he frames a more pressing leadership challenge: how school systems move from scattered experimentation to meaningful, measurable impact, without losing sight of trust, equity, privacy, and the human relationships at the center of learning.

Who’s in the conversation, and why it matters

Gilfus brings a rare blend of education and industry experience, as a founder of Blackboard Inc. to advising organizations on AI, cybersecurity, and digital transformation. Throughout the episode, he serves not just as a moderator, but as a guide, consistently connecting classroom realities to system-level implications.

Dr. Tom Ryan offers the long view. A longtime CIO, former CoSN Board Chair, and national leader in digital transformation, Ryan has lived through decades of technology cycles in schools. That experience shapes his skepticism of hype and his insistence on grounding AI adoption in instructional and organizational reality.

Narayan “Nandi” Nandigam, Vice President and Global Head of Services and Education at Infosys, adds a global and cross-industry perspective. His lens connects K–12 classrooms to enterprise systems, data foundations, and large-scale transformation efforts already underway around the world.

Why this moment feels different

Gilfus opens with a deceptively simple question: what feels different about AI compared to earlier waves of education technology?

Ryan’s answer is direct. This is not a traditional edtech rollout. “You can easily get into it on your own,” he explains, pointing to research showing that a majority of teachers are already using AI_attach AI—often without district-provided systems, training, or guidance. They are doing it for one reason: it saves time.

That reality changes the adoption equation. Districts are no longer deciding whether AI will enter their systems. It already has. The real question is whether leaders will shape its use or leave educators to navigate both opportunity and risk on their own.

Nandi reinforces the urgency from another angle. Generative AI lowers the barrier to entry by using natural language as the interface. When the technology speaks the same language as its users, adoption accelerates. That accessibility creates enormous potential for personalization and productivity, but also increases the need for clear leadership and guardrails.

“When the interface becomes natural language, AI stops feeling like a system you have to learn and starts feeling like a tool you can use.” Narayan “Nandi” Nandigam, Vice President and Global Head of Services and Education, Infosys

AI as a productivity accelerator, not “another edtech tool”

One of the strongest throughlines in the conversation is the rejection of AI as just another initiative layered onto already crowded agendas.

“Technology doesn’t causally impact increased academic performance,” Ryan says. “Teaching and learning do.” In his view, AI belongs in the productivity suite, not the edtech category. Its value lies in saving time, improving efficiency, and expanding access so educators can focus more deeply on instruction, relationships, and student engagement.

Gilfus builds on this idea by naming a familiar tension. Leaders often look for narrow proof points to justify AI adoption, while AI’s real impact is distributed across instruction, operations, and enterprise workflows. Measuring the right things, rather than expecting AI to single-handedly raise test scores, becomes critical.

Two examples that make the impact tangible

When Gilfus asks for concrete, tactical examples, Ryan offers two that resonate precisely because they are ordinary.

The first involves his daughter, a teacher preparing to teach 7th grade science for the first time. She was mapping out a pacing guide manually across months. Together, they used AI with state standards, the school’s instructional framework, and the district calendar to build a yearlong plan aligned to required outcomes. From there, they moved into unit plans and lesson plans that incorporated differentiation and established instructional frameworks.

“It wasn’t new work,” Ryan notes. “It was work she already had to do. It just made it easier.”

The second example is even more personal. A different teacher told Ryan, through tears, “I can finally be sick.” She had avoided taking time off because substitute plans in her subject area often fell apart. AI helped her create clear, usable lesson plans that a substitute could follow. The result wasn’t efficiency for efficiency’s sake. It was flexibility, resilience, and humanity in a demanding profession.

Leadership and values must guide adoption

As the conversation turns to ethics and responsibility, Gilfus presses both guests on what leaders should prioritize.

Ryan is unequivocal. “This isn’t an IT initiative.” It is a leadership initiative rooted in values. Education, he reminds listeners, is fundamentally relational. Districts must define what they value about great teaching and learning, then ask how AI enhances or diminishes those values.

“If our core value becomes making education cheaper at the expense of relationships or creativity,” Ryan warns, “we’re starting in the wrong place.”

Nandi complements that perspective with a practical responsible AI framework, emphasizing transparency and explainability, equity-first design, data privacy and consent, human-in-the-loop oversight, and ongoing professional learning. AI, he stresses, is meant to augment educators and administrators, not replace them.

Rigor, originality, and learner agency

Gilfus raises a concern many educators share: how do schools maintain rigor and originality when AI can generate content instantly?

Ryan reframes the issue with a line that lingers. “Hallucination is a feature, not a flaw.” Generative AI is designed to produce plausible language, not guaranteed truth. Treating it like a calculator is a mistake. Treating it like a thought partner can enhance creativity, reasoning, and critical thinking, as long as humans remain accountable.

“As AI takes on routine tasks, learning can shift toward reasoning, ethics, and conceptual understanding.”— Narayan “Nandi” Nandigam, Vice President and Global Head of Services and Education, Infosys

The implication is clear. Districts must explicitly teach verification, questioning, and judgment. AI does not remove the need for thinking. It raises expectations for it.

Nandi adds that this shift has the potential to move learning away from rote tasks and toward higher-order skills such as reasoning, ethics, and conceptual understanding. In that sense, AI forces education systems to clarify what they truly value and assess.

Infrastructure, governance, and the enterprise reality

Gilfus intentionally shifts the conversation to what sits underneath all of this: data, infrastructure, and governance.

Ryan emphasizes that AI adoption must be enterprise-wide, with executive leadership rather than sole IT ownership. He points to AI maturity models as a practical way for districts to assess current state, define future goals, and build roadmaps supported by evidence and measurable indicators.

Nandi expertly highlights common pitfalls: fragmented systems, poor integration, and uneven data quality. Drawing on his experience supporting large-scale transformations across industries, he notes that districts often underestimate the foundational data work required before AI can be used responsibly at scale. Without that foundation, even well-intentioned AI initiatives struggle to move beyond pilots.

He outlines key building blocks districts should consider, including governance structures, data foundations, responsible AI policies, change management, and flexible technology architectures that can evolve as tools and models change.

“AI outcomes are only as strong as the data foundation underneath them. That work is often underestimated, but it’s essential.”— Narayan “Nandi” Nandigam, Vice President and Global Head of Services and Education, Infosys

Stephen Gilfus’s closing challenge: start now, but start wisely

Gilfus closes the episode with a sense of urgency grounded in history. Referencing Moore’s Law, he asks listeners to consider the magnitude of technological change from 1963 to today, then project forward.

If change continues to accelerate, what happens to districts that wait?

His message is not reckless adoption. It is deliberate action. Start small. Secure data early. Experiment responsibly. Learn by doing. Because standing still is no longer a neutral choice.

From experimentation to impact

Taken together, Gilfus, Ryan, and Nandi make a compelling case: AI in Education is no longer theoretical. It is already reshaping how educators plan, teach, and manage systems. The real question is not whether AI will be used, but whether it will be guided by clear values, strong governance, and an unwavering commitment to human-centered learning.

Moving from experimentation to impact requires leadership more than technology. And that work, as this conversation makes clear, needs to begin now.

AI in K–12 education: what the numbers show

Survey data supports what the speakers describe. CoSN’s 2025 State of EdTech District Leadership report found that 80 percent of districts have generative AI initiatives under way or planned, and that 57 percent had placed AI guidelines in board-approved policy, with AI-enabled cyberattacks and student data privacy as leaders’ top concerns. An EdWeek Research Center survey found that half of teachers had received at least one AI training session by fall 2025, up from 13 percent in 2023.

AI in K–12 education at the core, not the edges

Stephen Gilfus has carried the episode’s argument further in his own writing. In a May 2026 field note on what he would build if he were rebuilding Blackboard today, he puts it in one line: “AI is the substrate, not the sidebar.” Tools that sit on the margin of existing systems, such as a chatbot on the navigation bar or a summarizer reading the same old tables, give a quick sense of progress, he argues, but each keeps its own memory and log, so a district cannot reconcile or audit what they do. He compares the two approaches to a campus power grid and a building full of extension cords: the cords are useful for a lamp and risky at scale, while the grid routes load, meters use and logs faults.

For district leaders, the practical test is ownership. Gilfus argues that procurement should ask whether a district can export, govern and audit what an AI tool does, not only its content, and he states the standard plainly: “If you cannot export it, route it, eval it, or govern it, you do not own it.” His advice for builders applies equally to districts: when time or money runs short, “reduce scope rather than compromise the architecture,” launching fewer uses with stronger governance rather than many that cannot be audited. On his own site he frames the stakes in leadership terms, writing that the forces reshaping how people live, learn and work will reward “the leaders who can govern and capitalize on that shift, rather than react to it.”

A district example: AI in the central office

Indianapolis Public Schools offers one example of the episode’s advice in practice. According to K-12 Dive, the district built internal AI tools on Google’s Gemini that score principal applications against its existing hiring rubric and draft standardized salary-offer emails using its compensation rubrics, without setting pay. It does not allow personal data in those tools, keeps people responsible for the hiring decisions, and reports savings of more than $300,000 for the 2026-27 school year. The approach follows the pattern the speakers describe: a narrow, time-saving use, clear rules on data, and human judgment where the decision matters.

A first year of AI in K–12 education

In his May 2026 field note, Gilfus lays out a quarter-by-quarter build plan for an AI-native learning platform. Adapted for a school district, the same sequence offers a practical first year:

• First quarter: choose one high-value use, such as feedback on student writing against an existing rubric, and run it with human review, logging and a set budget.
• Second quarter: extend that use end to end in two schools or courses, and agree what data the district owns and can export.
• Third quarter: introduce guided tutoring tied to specific learning goals, with a clear path for students to be handed to a teacher.
• Fourth quarter: widen the rollout only where the earlier steps can be audited, with records showing how each AI-supported decision was made.

At each stage the advice is the same: reduce scope rather than compromise governance, and expand only what the district can explain.

Questions and answers: AI in K–12 education

What does AI in K–12 education look like today?

Teachers are already using generative AI, often on their own and without district systems or training, mainly because it saves time on planning, lesson materials and routine tasks. With 80 percent of districts reporting generative AI initiatives under way or planned, the work in front of leaders is guiding that use rather than deciding whether to allow it.

How should a school district start implementing AI?

Start small and start deliberately. The speakers recommend beginning with a few well-chosen uses that save staff time, securing data early, experimenting responsibly and learning from what happens, with human review built in from the first day. Stephen Gilfus adds that when time or money runs short, it is better to reduce scope than to launch tools that cannot be governed or audited.

Who should lead AI in a school district?

District leadership, not the IT department alone. Dr. Tom Ryan describes AI as “not an IT initiative” but a leadership initiative rooted in values, and argues that adoption must be enterprise-wide with executive leadership. Technology teams remain essential partners for data, security and integration.

What does a responsible AI framework for schools include?

Narayan Nandigam outlines five parts: transparency and explainability, equity-first design, data privacy and consent, human-in-the-loop oversight, and ongoing professional learning. Many districts are putting these commitments in writing; CoSN found that 57 percent had placed AI guidelines in board-approved policy.

How can schools protect rigor and originality when students use AI?

By teaching verification, questioning and judgment explicitly. Generative AI produces plausible language, not guaranteed truth, so Ryan cautions against treating it like a calculator. Used as a thought partner, with people accountable for the final work, it can support reasoning and critical thinking, and Nandigam argues it can shift learning toward higher-order skills.

What is an AI maturity model for schools?

An AI maturity model is a structured way for a district to assess where it stands today, define where it wants to be and build a roadmap with evidence and measurable indicators. Ryan points to maturity models as a practical tool for moving beyond pilots. The CGCS and CoSN Gen AI Readiness Checklist listed below covers similar ground across leadership, policy, data and implementation.

Does AI improve student achievement on its own?

Not by itself. “Technology doesn’t causally impact increased academic performance,” Ryan says. “Teaching and learning do.” The speakers suggest measuring AI where its value appears, such as staff time recovered, planning quality and continuity for students when a teacher is out, rather than expecting a single tool to raise test scores.

For a step-by-step plan, see edCircuit’s AI adoption roadmap for K–12 schools, and for benefits, risks and policy across all levels, the AI in education guide.

Helpful Resources

• K–12 Generative Artificial Intelligence (Gen AI) Readiness Checklist
  Developed by the Council of the Great City Schools (CGCS) and the Consortium for School Networking (CoSN), this checklist helps districts assess readiness across leadership, policy, data, and implementation

This CoSN podcast episode was sponsored by Infosys.

Subscribe to edCircuit to stay up to date on all of our shows, podcasts, news, and thought leadership articles.