Five Things I Brought Home from NACAC

I'm back from NACAC, the national conference for college admission professionals. I went to see how higher ed enrollment teams are working with AI, and what's worth bringing back to K-12. Higher ed tends to run a few years ahead of us on the office side of this work, and on the family side it turns out we aren't behind at all. Here are five things I brought home.

1. AI stopped being the news and became the job

In May, after ATLIS, I wrote that AI had already arrived in schools. At NACAC, it felt like it had settled in. The frantic energy of "AI from every vendor" has given way to a better question: how are you actually using it? AI is still the dominant topic, but the conversation sounds more like the normal evolution of a new technology than a revolution.

2. Chatbots are everywhere. Agents are still the edge

In the sessions and throughout the exhibit hall, much of the discussion focused on colleges' use of chatbots and AI assistants to answer families' questions. I heard a lot about agentic AI, but saw only a few sophisticated working examples. My sense is that many people understand agents as a concept, but are still hazy on what one would do in their own workday.

The agents people were talking about in the hallways were consumer ones: Meta's Muse and ChatGPT's new dots. Both can take actions on your behalf, and Muse in particular is built for everyday users. The phrase I kept hearing was "It's the AI your parents can use." Meta has earned some distrust in how it handles data, and I understand that. I still think a careful, limited look at Muse is worthwhile. Tools like this could change how quickly a broader group of people understand what an agent is.

3. In some places, adoption is going backward, and that's healthy

People who leaned in early are pulling back where AI didn't work for them. Some uses that were popular a year ago, like drafting recommendation letters, are now a line many counselors won't cross. One speaker described pulling back after realizing he'd been careless about where he drew his human line.

4. Permission still lags behind use

Most of the AI sessions I attended began by polling the room: who's using AI, and how much? Nearly every hand went up. Far fewer people had heard their leadership say it was okay. Everyone wants a policy, but few are sure what belongs in it, especially when the technology keeps changing. Given how fast the ground is moving, that's understandable. It also means silence from leadership gets read as either "no" or "anything goes."

5. Families ask AI about your school before they ask you

This was my biggest lightbulb moment, and it came up in nearly every AI session I attended. The speakers were different and so was their data, but the picture was the same every time: the families you hope to meet this year are asking AI about your school before they ever reach your website.

The numbers first. Ben Neely of Revolution Prep opened the preconference with three years of his own student survey: the share of students using AI to research colleges went from about one in nine to about half in two years. His one-word summary of what AI is replacing: Google. At another session, EAB (formerly the Education Advisory Board, a research and consulting firm) reported that more than half of parents of high schoolers now use AI to research colleges as well. The habit doesn't switch off when the school is a kindergarten or a ninth grade.

How are families asking? Madeleine Rhyneer of EAB read out the prompts students actually type: "I have a 3.6 GPA, what schools are competitive for me?" "My family makes about this much, what schools are a good financial fit?" Few of them type "best colleges." They describe themselves. The K-12 version is a parent describing a child.

And where do they land? Her co-presenter, Dee Voss, who leads enrollment at The New School, has measured it. Students read the AI summary and skip the homepage. Someone asking about the best fashion program in New York goes straight to the fashion page, so every program page on your website is now a front door. The website's job has changed. It no longer introduces your school. It confirms or breaks what the family already heard. And the inaccuracies families run into, Madeleine said, mostly come from the school's own pages.

In July, I suggested a mirror test for K-12 enrollment leaders: ask a few AI models what a parent in your area would hear about schools like yours. NACAC turned that hunch into a to-do list, with two refinements.

First, test the version of AI families actually use. A figure that came up in the room: about 98% of U.S. households don't pay for AI. The answer a parent sees comes from a free version, often logged out, and it isn't the answer your team sees on the paid account at work.

Second, ask a question you'd rarely ask yourself. Madeleine suggested asking AI what kind of family would not be happy at your school. The answer shows how the internet sees you, and it can sound a lot like the hesitations families bring with them on a tour.

The phrase I'll remember

The phrase I'll remember from this conference is "healthy skeptic." I heard it most from the people using AI the most. Nearly everyone admitted to being worried (how could you not be?). The "healthy" part means staying engaged. It's hard to shape good policy around a tool you don't understand, and the careful voices are already leading to more thoughtful controls.

The mirror test is what staying engaged looks like for an enrollment office. I came home wanting to understand it better than a hunch allows. Data and research, with AI's help, can show us what happens behind the scenes when a parent types a prompt. I've started that research, with the goal of giving school leaders concrete examples relevant to K-12 schools and the families they serve. I'll share what I find here and on LinkedIn over the coming weeks.