Vice Provost on the $100M Guessing Game in Higher Ed

Karthik:

“You started in broadcast communication and ended up running enrollment for major universities. How did that happen, and does your communication background shape how you approach enrollment?”

Steve:

I went to college wanting to be a sportscaster. I had an internship at CNN Sports in Atlanta in 1991 and learned what happens behind the scenes. That industry was more cutthroat and self-serving than I was cut out for. I am more collaborative, the lift-everybody-up type. In college I was a resident assistant, helping younger students adjust and stay on track, and I was a cheerleader, so I had a very rah-rah relationship with my institution. After I graduated I went to work at the same school in recruitment, driving back and forth across Wisconsin and visiting 200-some high schools in a season. I just stayed with it.

Karthik:

“You describe successful enrollment management as a relationship between the student and the university. What does that look like in practice today?”

Steve:

Ask a few enrollment managers and you will get different answers. Mine goes back to my education: I studied communication for my master’s and PhD too. I believe the relationship between a student and the institution is interpersonal. When people talk about their school, they talk about it like a character in the story. Every exchange with the institution is with a person in that moment, and each of us takes a turn holding the role of “school” before handing it to someone else. Housing, billing, faculty, advisors, the president, even roommates all take turns playing that role. Enrollment management started with everyone acknowledging their part in that relationship so students feel a sense of belonging and choose to stay. When the institution stops playing its part, the student feels lost or disconnected, and there is nothing to protect. It is like any relationship: some you can pick up years later, many you cannot. Schools that recognize this and integrate it across the whole experience do well. Schools that do not usually struggle with retention.

Karthik:

“What has changed in the expectations of students and families today compared with when you started?”

Steve:

I started in 1993, when there was no internet. If you wanted to talk to someone, you called or drove to their hometown. Students often heard about your school for the first time when you walked into their high school. That epiphany does not happen anymore. Students have been on our websites and are receiving our messages long before we meet them. Higher ed buys prospect names, and I buy names of high school freshmen and sophomores, so I may be interacting with a student for years before we ever meet face to face. The recruiter used to establish the relationship. Now we enter conversations that are already established. We are no longer convincing anyone to consider college. We are differentiating our school from someone else’s. We are essentially playing defense.

The whole marketplace is different too. After World War II, the GI Bill and student loans tied higher ed to the federal government, partly to avoid the post-war recession that followed World War I. It shaped the next 75 years: go to college, get a good job, move into the middle class. Barack Obama said that in three State of the Union addresses. Today that relationship is strained, and the perception that students do not really need college has never been as strong. Whether that comes from AI, social media, or something else, people need to be convinced again that college is worth their time.

Karthik:

“Financial aid has been turbulent the last few years. How much of your team’s energy goes into absorbing federal aid changes and regulatory shifts versus regular enrollment work?”

Steve:

The short answer is a great deal, though not equally across all teams. In higher ed, if something does not involve you, you tend not to pay attention. When the financial aid application was completely changed for fall 2024, the fact that the FAFSA was not ready was an enormous inhibitor to recruitment nationwide, because 70 to 80 percent or more of students at many schools need aid and cannot get it without that form. The One Big Beautiful Bill that passed last year was largely focused on graduate borrowing, especially health-related graduate programs. I have only been at my institution six months, but I have spent a lot of time on it. Right now I am at Texas Tech University’s Health Science Center, so those changes affect easily 80 percent of my total enrollment. A school without graduate health programs would feel it differently. Since the 2008 housing crash, higher ed has paid far more attention to the economy and the government’s relationship to higher ed than it ever had.

Karthik:

“For someone watching, what are the stages from a student raising a hand to sitting in a classroom?”

Steve:

Schools start in slightly different places, but they converge quickly. Some start with a suspect, others with a prospect. Then comes applicant, then admitted student, then matriculation, which is roughly signing up for classes. It is not over until the student is actually sitting in class, and enrolled headcount is usually measured at census, around the 14th day. The whole process can start two years or more in advance. The difference between suspect and prospect is whether the institution knows who you are. At previous institutions I might buy 300,000 or 400,000 names. I do not have that many interested people, so when I mass-market to them, they are suspects. When someone responds and says, yes, send me that, I am interested, they become a prospect, and different communication plans target them from there.

Karthik:

“Where in the funnel is the biggest leakage? If you could fix one stage, which would it be?”

Steve:

This may sound critical of today’s college-bound decision maker. Picture yourself in a clothing store in front of a wall of jeans: every color, style, brand, and size. If you know what you want, you go straight to it. If you have not thought about it, you stand there a long time. That is decision theory. Higher ed does not have trouble getting applications. They are easy to submit, and admissions decisions are easy to get. The leak is in preparation to decide. If you applied to 10 schools, you eventually have to take nine off your list, but students do not actively remove the ones they have lost interest in. They pay deposits at multiple institutions and attend orientations at more than one because they are still unsure. So everyone waits until May 1, then there is a landslide of decisions, followed by substantial melt. I have been at campuses where melt was 10 to 15 percent. In a class of 2,000, that is 300 people who paid a deposit and did not come.

The old funnel worked because students eliminated schools by not applying. Now we get applications from everyone, and students stay in contention until the end. They stop reading our messages, but it is hard to take financial aid back because someone did not answer an email. So we commit resources, time, and seats to people who probably already know they are not coming.

Karthik:

“Is summer melt the most frustrating problem?”

Steve:

It is the most visible, but I do not think it is the most problematic. The bigger problem is the number of applications. Schools say the application is free, just hit submit, and tools like the Common Application let you check more boxes and send it everywhere. We have made it so easy to apply that we created our own confusion. If you are genuinely interested in 20 schools, I would argue you are not interested in any. Every year there is a news story about a student admitted to every Ivy League school. They are all so different, in different places. If you apply to all of them, you do not know where you want to go. After all, the Ivy League is an athletic conference. We have replaced decision-making with more-is-better. Students could save themselves a lot of time by letting their 20th school off the hook.

Karthik:

“What is the most manual, unglamorous operational work in enrollment that you would like to automate?”

Steve:

Probably transfer credit evaluation, which is also among the most important. Transfer credit should be largely predictable. Courses at a known institution should be known entities. Yet some schools evaluate credit only after a student applies, is admitted, and pays a deposit, while others tell students how many credits will transfer before they even apply. Every school thinks its process is unique, but it ends up exactly the same nearly everywhere. Nobody takes an English class and pretends it is a math class. The only people who do not know that are the people who work there. Transfer students know which schools treat them efficiently. Waiting until after matriculation just adds weeks or months to a process that offers nothing new.

Karthik:

“How fast do teams respond to inquiries across channels, and is that pace good enough?”

Steve:

Having been here only six months, I will answer for the industry instead. Speed matters. You have heard that speed kills; in higher ed, the lack of it is what kills. Every process is better when it is faster. International students are being told to plan their US arrival more than 365 days in advance, so every day we take to evaluate a document is a day they lose to secure a visa. With transfer credit, if school A has already evaluated your credits and school B asks you to collect every syllabus from every old course, the competitor is done. The only person who knows it cost you an enrollment is the student who already removed you. I have seen a student with 54 credits, that is 18 courses, asked to call every old institution for syllabi, and they did not finish the application. Syllabi live in LMS platforms students can no longer log into. We stopped handing out paper syllabi 20 years ago, but our processes have not caught up. If a student already has an answer elsewhere, that is good enough for them.

Karthik:

“Would connecting systems like the CRM, financial aid, and communications tools speed things up?”

Steve:

Yes, but how often do they talk to each other? Not much. Higher ed treats the source of truth as a very important principle, and I think it has become dated. There is a strong aversion to keeping data from the student information system, Banner or PeopleSoft, in another system like Slate. I do not want to manage curriculum in the CRM. But for a CRM to follow best practice, I need students’ registration recorded there. If a student has not registered by the midpoint, I should message them, because they should register, and because if they are not going to, I need to follow up. That is different from a generic blast: “Dear student, we notice you haven’t registered. Love, Steve.”

In communication theory, all communication is relationship-serving: the status of our relationship is embedded in the words. Amazon’s “people like you also bought” messages show you they see who you are. Higher ed needs that. “Tommy, you usually register by now and you haven’t” tells Tommy that I know him, I know his pattern, I noticed it is missing, I expect a problem, and I want to help. A general message cannot do those five things. The systems do not all need to replicate the student information system, but they must share the data points that let those tools convey the relationship. Students rebuild their belonging when they notice that we noticed they were missing.

Karthik:

“With ChatGPT and answer engines, how can AI be used effectively in enrollment? Are institutions prepared for these channels?”

Steve:

Every higher-ed-adjacent company is ringing our doorbell about AI, and if it makes my life better, great. The dangerous part is “and then I will talk to your students so you do not have to.” Higher ed has been reluctant there, because of our history as a first-person relationship with students. Some students would not mind talking with an agent, but not all, so we will always need a human version running alongside. We already send different messages to Gen X parents and Gen Z students. AI may know what to say without understanding that the person does not want to discuss personal matters with an agent. I would not want a family’s financial aid information in an AI agent, but I can see an agent contacting a student to set up a financial aid meeting. Engagement, yes. Replacement, probably not. There are times we must be empathetic and say no at the same time, and I do not think anyone wants AI doing that for the institution.

Schools across the country are also receiving an inordinate number of what I believe are AI-generated fraudulent applications, including fraudulent financial aid applications that get paid out. Proving we are talking to real students matters for compliance and accreditation. Families feel the same way. They do not want to commit to a $50,000 or $75,000 tuition bill and realize they never spoke to anyone. Picture families screaming at a phone menu the way we all do about a credit card bill. This is arguably one of the biggest expenses a family will ever have, and we have to honor that with a human touch.

Karthik:

“Besides financial aid, are there other areas where AI cannot replace people, at least with today’s technology?”

Steve:

I have anxiety about using it to identify mental health risk. An AI agent can only interpret what is shared, and tone is hard to read without caps or exclamation points. When people are at their most needy, it is important that they experience a human on the other side. Schools with 50,000 to 100,000 students may have to use those tools, but for the 95 percent of us serving populations we can adequately handle, that clarity matters. Businesses say with pride, “you will talk to a real person.” If schools say that, it says a lot about our character and identity.

Karthik:

“If a technology team like ours wanted to build tools for enrollment offices, what should we build?”

Steve:

That is a question tech firms do not ask enough. People call and say, here is the AI tool I built, would this make your life better? I wish they would ask what tool would make my life better, and then build it. One of our biggest challenges is mass changes in the student information system across thousands of records. IT can run a script, but often without the ID of the person making the change. For financial aid notes, record updates, or correcting mass errors, I can write what I need in an Excel file, but in the system I may have to touch three, five, or seven pages per record. The same goes for transfer credit: store the day’s evaluations centrally and push them all in, instead of typing them one course at a time.

Here is an example. Say an institution coded its campuses by name, then renamed a campus eight years later. After 20 years in Banner, you have legacy codes that no longer mean anything. To fix one of those for 5,400 students without a batch function, someone updates 5,400 records by hand. Each change takes 15 or 20 seconds, but between calls and email it can take a month. Across 10 institutions, my internal challenges have been a bigger concern than working with students, which we actually do pretty well. If I had an AI tool I could deploy and set conditions for, one that goes and gets the registration date and populates it, I could manage that faster than waiting for IT to build an integration that only works one way and can break. That is far more useful than trying to replace me in an exchange with a student.

Karthik:

“What advice would you give someone stepping into their first enrollment leadership role this year?”

Steve:

A lot of campus decisions are made on anecdotal evidence and long-term memories. That data was probably true once upon a time. Go back and do your own research to find out what is true, so you can make data-informed decisions when you need to. The only time to trust anecdotal evidence is at a retirement party when people are telling jokes about the old days. There is no value to it when you are dealing with $100 million operations, where decisions hinge on knowing how the variables will move. Our institutions are responsible for more money each year than the banks in the same town. These are financial enterprises, and making decisions on anecdotes is tantamount to guessing. In 2026 and beyond, guessing is not the way to run operations this size.