AI ROI

Four business professionals gather around a conference table as one person presents a printed report with a circular chart beside an open laptop.

Most districts didn’t adopt artificial intelligence carelessly. You had a district improvement plan. You had specific content area goals. When you brought in an AI tool, you likely had a real reason for it, maybe it was supposed to support a literacy initiative, ease the load on your special education staff, or help teachers personalize instruction faster. The plan wasn’t missing.

Here’s a question worth sitting with. If a school board member or a reporter asked you right now what specifically changed because of the AI tools your district already has, could you answer? Most leaders can’t, not because the tools failed, but because nobody wrote down what “changed” was supposed to mean before the tool showed up. That gap is the actual problem this post is about.

What’s usually missing is the connective tissue: a clear, written answer to how a specific AI tool supports a specific goal, and whether it actually can. Many districts have responded to that uncertainty in a reasonable way, by keeping AI use teacher-facing for now and away from direct student contact, given legitimate concerns about safety and student data. That’s not indecision. It’s a sensible starting point. But it changes what you should be measuring. If students never touch the tool directly, your return on investment (ROI) question isn’t about learning gains showing up on a test. It’s about capacity: did this tool free up time or attention for your teachers, and did that reclaimed capacity actually go toward the goal it was meant to serve?

Why this is harder than it looks

A new curriculum or intervention program usually comes with its own success metric built in. You get benchmark scores, fidelity checklists, and a vendor who tells you exactly what “working” looks like. AI tools rarely come with that. Worse, many districts haven’t worked out the causal chain between “this AI tool is available to our staff” and “this specific line on our district improvement plan.” That chain has to be built on purpose. It doesn’t arrive in the box, and for a lot of districts, it didn’t arrive with a box at all.

The AI That Came with the Platform

A teacher works alone on a laptop in a sunlit classroom, surrounded by lesson plans, notebooks, sticky notes, student papers, and classroom supplies.

Not every AI tool in your district was purchased as a separate product. Depending on your platform edition, licensing, and administrative settings, AI features such as Gemini or Copilot Chat may already be available within systems your district licenses. In other cases, a tool received a general nod of approval—“this can help with lesson planning”—without ever being tied to a specific goal. None of this is necessarily a personal or unofficial use question. The tool may be district controlled and officially available while still sitting there largely unexamined.

This category is genuinely harder to audit than a tool you bought on purpose. A purchased tool eventually comes up for budget review, and that renewal conversation forces someone to finally ask if it’s working. A bundled tool usually renews automatically as part of a much larger platform contract, so it can go years without anyone isolating it as its own line item or asking what it’s actually for.

The fix isn’t complicated, but it does have to be deliberate. Schedule a periodic review, once a semester is reasonable, of every AI feature that’s turned on by default across your platform licenses. For each one, make an explicit choice: assign it a specific purpose and start measuring it, restrict or turn it off if it isn’t earning its place, or leave it available with clear guidance while being honest that it isn’t being measured yet. Any of those three is a defensible answer. Leaving it on with no decision at all is the one that isn’t.

Consider a middle school that purchased an AI tool specifically to support a literacy goal on its district improvement plan. Leadership deployed it teacher-facing only, for lesson planning and feedback drafting, which made sense given the safety concerns at the time. Six months later, nobody can say whether it moved the needle on that literacy goal, not because the tool failed, but because no one defined, before rollout, what “supporting the goal” would actually look like. Was success measured in hours of prep time reclaimed? Faster turnaround on writing feedback? Something else entirely? Without that target, the campus has no way to know if the investment is paying off, even if it is working well right now.

Setting your own target

Even though the semester has already started, it’s not too late to name two or three things you expect to change because of a tool, whether you purchased it deliberately or it came bundled into a platform you already had. Be specific. Instead of “improve literacy instruction,” try “reduce the time teachers spend drafting written feedback on student essays by half” or “increase the number of small-group intervention sessions a teacher can run each week.” Write these down, share them with your campus administrators and instructional coaches, and check them explicitly rather than relying on a general satisfaction survey. A survey tells you how people feel about a tool. It doesn’t tell you whether the specific goal you set out to support actually moved.

The AI you didn’t purchase

A woman works on a laptop at a dining table at night with an open planner, pen, coffee mug, sticky notes, and smartphone nearby.

Here’s a layer many districts haven’t fully accounted for. Alongside any officially provided tool, bundled or deliberately purchased, your teachers and staff are almost certainly already using free or personally purchased AI tools, on the school network during the day or at home in the evening. They’re using them for lesson planning, parent communication, feedback drafting, and more. This is happening whether or not you’ve measured it, and it deserves an honest look rather than being ignored.

Two things follow. First, find out what’s already happening informally. You may discover both hidden value, teachers who’ve quietly found real efficiency gains, and hidden risk, particularly around student data ending up somewhere it shouldn’t. Second, give your teachers a simple, personal way to gauge value for themselves. Not a district form. Just a private gut check: did this save me real time this week, and would I miss it if it disappeared tomorrow? That question alone helps a teacher separate a tool that’s genuinely earning its place from one that’s just become a habit.

Whether a tool is bundled, deliberately procured, or personally chosen, the same rule applies. Never enter personally identifiable student information into an AI tool that hasn’t been vetted and approved by your district.

The equity layer you can’t see from the district office

If you only measure AI’s impact at the aggregate district level, you risk missing something important. One campus might show strong results while another falls further behind, and averaging the two together can hide that gap entirely. This applies to deliberately purchased tools, where some campuses may have received more training or more consistent access, and it applies just as much to bundled tools and personal use. A teacher with a paid personal subscription may be getting meaningfully more value than a colleague relying on a free tier, a gap your district didn’t create and might not even see, since it’s happening on personal devices and personal accounts. Disaggregate your data by campus wherever you can, and ask specifically whether your rollout, official or otherwise, is widening any existing gaps.

Budget reality

Split-screen image of two educators working on laptops in separate classrooms, surrounded by books, lesson materials, binders, and educational displays.

A lot of AI spending over the past few years leaned on temporary funding, grants, one-time allocations, or ESSER dollars. That ESSER money is gone now. The final spending extensions expired in March of this year, and bundled tools carry a related risk of their own, since they can quietly disappear or change scope whenever the parent platform contract is renegotiated. If an AI tool in your district was ever propped up by ESSER funding, you’re not waiting for a funding cliff anymore. You’re standing at the bottom of one, deciding right now whether that tool earns a permanent, deliberate place in your general fund budget. Setting a clear target today gives you the evidence you’ll need to make that case, one way or the other, in a conversation that’s already happening rather than one still ahead of you.

If you’re heading into a tighter budget year and need help deciding what to keep, improve, or cut, contact Dr. Bruce Ellis at bellis@tcea.org to learn more about TCEA’s AI Investment Audit.

The audit gives districts a clear picture of where AI is being used, including tools that were bundled or pre-approved, where adoption is falling short, and which investments are earning their place. That means your next decision can be based on evidence instead of guesswork.

What to do this week

The next step looks a little different depending on your role.

Vertical workflow diagram showing a professional checklist feeding into an AI processor, which then connects to a laptop displaying rising analytics and a settings gear.

You already had a plan when AI entered your district. Now give yourself a way to know if it’s working.

Does Your AI Support Your District Improvement Plan? A Practical AI ROI Check