AI in Higher Ed

AI is exposing your org chart.

Most AI projects in higher education work exactly as sold and change nothing, because the tool sits inside a structure built to hand students off between departments. A chatbot that answers at 11 p.m. still loses the student if nobody opens the inquiry queue until Monday. I've run AI at institutional scale, and the constraint has never once been the model.

What I've built

At Forsyth Technical Community College we deployed a chatbot called BlazeBot to handle the questions a person shouldn't have to answer twice. It works. Advisers got hours back, and students get answers when they're awake and looking, which is rarely during business hours.

The part that took longer was everything around it. Deciding who responds when the bot escalates. Deciding what counts as answered. Agreeing that a student who asks a question at midnight has started a relationship with the college. None of that is a technology decision, and all of it determines whether the technology matters.

I'll tell you the parts that didn't land, because that's the useful half. We've bought things that solved a problem we didn't have. We've automated a process that should have been deleted.

One more thing I've stopped arguing about. The question of whether AI belongs in student communication is settled by students, not by us. They're already using it to decide where to apply. A college that opts out of the conversation doesn't remove itself from the answer, it just loses the ability to correct what's in it.

What I argue

The escalation path is the product

The demo always works. What breaks is the share the bot can't answer, because that lands on a person who already has a full queue. Before the contract is signed, somebody should be able to say out loud who picks up an escalated question at 4:45 on a Friday, and what happens to it over the weekend.

Speed only counts if someone is behind it

An instant automated reply followed by four days of nothing is worse than a slow human answer. The student learns something either way, and what she learns from the fast empty one is that the college is a machine that noticed her and then didn't care.

Augmentation is a scheduling decision

Freeing up adviser time means nothing unless somebody decides what the time is for. If the hours the chatbot gives back just absorb a larger caseload, the college bought efficiency and called it care. Deciding to spend that time on the conversations a person has to have is a management choice, and it has to be made on purpose.

Watch what students are asking the model, not you

Prospective students are running their research through a system you can't see and can't edit. By the time they reach your site, they've already been told something about your college. Nobody in higher ed has a clean answer for that yet. Checking what the models say about you is at least a start.

Pilot where the failure is cheap

The instinct is to put AI in front of prospective students first, because that's where the volume is and the demo looked good there. It's also where a wrong answer costs the most. Start somewhere internal and boring, let the team learn what the tool is bad at, then move it toward students once you know.

Writing on this

Community College Marketing Has a Discovery Problem

AI search changed where students start. The college that answers the question well is not always the one closest to them.

Your AI Strategy Won't Work Because Your Org Chart Won't Let It

Adoption is climbing across higher education and outcomes aren't following. The fragmentation was always there. AI made it impossible to ignore.

The Question behind the Submit Button

Students are using AI to do research colleges used to do for them, and they've already decided what a slow reply means.

Where this shows up

I speak on this as "AI That Reduces Friction, Not Humanity" and "Building AI-Ready Institutions," usually for rooms that have already been sold the tools and want to know what happens after procurement. The consulting version sits under AI and digital infrastructure, and it starts with escalation paths and the org chart rather than the vendor demo.

The colleges that get value out of this decade won't be the ones that bought first.

Related subjects

Student experience

The structure the tool gets dropped into, not the model.

Enrollment marketing

Where AI search changed how students find a college at all.

Work together