Operating Philosophy: People
Why AI Adoption in Higher Education Starts With People
AI adoption depends on more than technical capability. Culture, faculty judgment, incentives, and institutional relationships determine whether a product meaningfully improves how people work.
People is the first pillar of my People, Business, Technology framework. It asks who is affected, what they need, and what would make them adopt or resist a change.
Most institutions I talk with can name an AI tool it has bought or piloted. Far fewer can name one that changed how teaching or assessment actually works. The distance between those two lists is rarely a technology problem.
People define the aspiration. Culture, incentives, and relationships decide whether change actually happens. That is why I weigh every decision in the same order: people, then business, then technology. Technology introduced without understanding people usually becomes technology people tolerate, not technology that changes anything.
The delivery risk is people
Before I worked in AI, I spent four years on large SAP S/4HANA transformations. The software was enormous and complicated, but it was not the hardest part. The hardest part was coordinating hundreds of subject-matter experts across companies, countries, and time zones, each with a day job and a good reason to keep doing things the old way.
I ran live Scrum sessions, recorded training, and supported more than 200 client experts through those rollouts. The lesson carried straight into education. A system tends to succeed when the people expected to use it can see their own work getting easier. Until they can, every feature is a request for extra effort.
Culture: faculty are not an obstacle to route around
Higher education has a culture built on expertise and judgment, and that is a strength. Faculty decide what counts as learning in their course. An AI tool that quietly takes that decision away will meet resistance, and it should.
Resistance is not always resistance to technology. Sometimes it is resistance to losing ownership of an academic decision. Sometimes it is one more workflow added to a full semester, or a tool that has not yet shown it is worth the time. Each is a fair objection, and each calls for a different response.
The tools that earn adoption do the opposite. They take on the repetitive work that leads up to a decision and leave the decision itself with faculty. Drafting feedback a professor reviews is a different proposition from issuing a grade a professor has to defend. The first respects the culture. The second asks faculty to give up the thing that makes them faculty.
Incentives: students respond to what we measure
Students respond to incentives. If a grade rewards a finished product, and AI makes finished products cheap, students will use AI to produce them. Calling that a character flaw misses the point. It is an incentive we built.
Detection tools try to fix this by raising the penalty. That approach has two problems. The tools are not reliable enough to support an accusation on their own, and the threat does nothing to change what the assignment rewards.
When we begin evaluating a student's reasoning, revisions, and understanding alongside the final submission, we change what the assignment rewards. AI can still contribute, but simply generating the answer becomes less valuable, because the student needs to show how they arrived at it.
The question I keep coming back to is simple: when does support become substitution? Students can't answer that for themselves if our assignments never ask it.
Relationships: trust is the real infrastructure
Every AI policy runs on relationships: between faculty and students, between departments and IT, between an institution and its vendors. A false accusation from a detector damages the first. A tool pushed onto faculty without their input damages the second. A vendor that overpromises damages the third.
I have seen the reverse, too. One of the most vocal early skeptics of the AI suite I work on, at one of our institutional partners, became one of its strongest advocates. What changed their mind was not a sales pitch. It was roadmap and interface decisions I directed to fix real higher education workflow problems.
Skeptics are often the people who understand the work best. Their objections are free product research, if you are willing to listen to them.
What this means if you lead AI adoption
- Start with the workflow, not the feature. Ask faculty what takes their time and what they refuse to give up, before anyone demos anything.
- Find your skeptics early and give them the hardest cases. If it holds up with them, it will usually hold up with everyone else.
- Keep a person accountable for every decision AI informs, and make that visible to students.
- Measure adoption by what changed in practice, not by licenses issued or logins counted.
When technology falls short in higher education, it is usually not because it can't do the job. More often, nobody asked the people doing the job what they needed, what they feared, and what they were measured on.
Understanding people tells us what should change. The next question is how to define, measure, and sustain an outcome worth pursuing.
Part of my People, Business, Technology series. Next: How to Measure the Real Value of AI in Education. Also read: Technology Should Solve the Problem, Not Define It.