Short answer: AI agents in higher education admissions are assistants that answer applicants around the clock, remind each student about the steps they still owe, and pass hard questions to staff. The best evidence comes from Georgia State University, where a randomized trial of its "Pounce" chatbot made committed students 3.3 percentage points more likely to enroll on time, a 21% reduction in summer melt. Adawaty's AI Admissions Desk brings the same model to your school: live in 10 days, $990 + $149/mo.

Key facts

  • Summer melt, where admitted students who plan to enroll never show up, affects an estimated 10–20% of college-intending students each year, according to Page and Gehlbach.

  • In Georgia State's 2016 trial, 7,489 admitted students were randomly split. Committed students who got Pounce were 3.3 percentage points more likely to enroll on time.

  • Only 13.5% of students in the trial sent a message the chatbot couldn't answer and had to pass to staff.

  • The chatbot platform cost an estimated $7–$15 per student per year, against $100–$200 for earlier counselor-outreach programs.

  • Two later randomized trials found course chatbots raised the chance of a B or higher by 5–6 percentage points.

  • In EDUCAUSE's 2025 study, 57% of respondents said their institution sees AI as a strategic priority, but only 22% had an institution-wide AI strategy.

An admitted student is not an enrolled student. Between the offer and the first day, there are forms, fees, documents, deadlines and a lot of silence. Most institutions lose students in that silence, and never find out why. Georgia State decided to measure it.

What happened at Georgia State?

In 2016, Georgia State University worked with the education start-up AdmitHub to launch Pounce, an AI text-message assistant named after the university mascot. Researchers Lindsay Page and Hunter Gehlbach ran it as a field experiment and published the results in AERA Open in 2017.

  • The setup: 3,745 admitted students got Pounce. 3,744 got the usual process.

  • The data link: Pounce was connected to the university's student information system. It only messaged a student about a task that was still open, such as a missing transcript or an unfiled financial aid form.

  • The knowledge: admissions staff seeded about 250 frequently asked questions. By the end, the knowledge base had grown to more than 1,000 answers.

  • The handoff: questions Pounce couldn't answer went by email to staff, and the staff reply went back to the student by text.

The result: among students who had already committed to Georgia State, those who got Pounce were 3.3 percentage points more likely to enroll on time. The authors say that equals a 21% reduction in summer melt. Committed students also did better on financial aid steps. For example, they were 6 to 7 percentage points more likely to accept a loan they had been offered.

The authors estimate that across a full intake of about 3,500 students, the effect equals roughly 116 extra students starting in the fall.

Why did it work when so many chatbots don't?

The study design answers this, and it is the part most schools skip:

  • It knew each student's status. Because it was linked to student records, Pounce only nudged students about what they still owed. No generic blasts.

  • It went where students already were. Text messages, not a portal they had to log in to. About 85% of students replied at least once.

  • It answered instantly, at any hour. Students got answers on their own schedule, 24 hours a day.

  • Humans stayed in the loop. Hard questions went to staff, and their answers made the bot smarter.

A 2023 follow-up by Page, Meyer, Lee and Gehlbach tested chatbot outreach again at a two-year and a four-year college. Its lesson is honest and useful. Outreach worked best for discrete, time-sensitive administrative tasks, like filing aid forms or registering for courses, targeted only at the students for whom they were relevant. Messages nudging students toward less urgent supports, like meeting an advisor, had little to no effect.

Can AI help after enrollment, too?

Yes, with the same rules. In 2021, Georgia State added a course-level chatbot to its large online American Government class. Students got reminders about assignments and exams, plus a way to text questions to the instructor.

  • A paper covering two randomized trials, in political science and economics, found the chatbot raised the likelihood of a B or higher by 5–6 percentage points and reduced course drops.

  • Georgia State reported that first-generation students who got the messages earned final grades about 11 points higher than similar students who didn't.

Timothy Renick, founding executive director of Georgia State's National Institute for Student Success, called these "exceptional results from a fairly light-touch initiative."

Where do most institutions stand on AI right now?

Interested, but not organized. The 2025 EDUCAUSE AI Landscape Study found:

  • 57% of respondents said their institution views AI as a strategic priority, up from 49% the year before.

  • Only 22% said their institution had an institution-wide AI strategy. 55% said strategy was happening in pockets.

  • Just 2% reported new funding sources for AI projects.

That gap is the opportunity. Admissions and student services have clear tasks, clear deadlines and clear data. That makes them one of the easier places to show results without a campus-wide AI program.

What is coming in 2026–2027?

This section is outlook, not established fact:

  • From chatbots to agents. Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024. In admissions, our view is that this means assistants that don't just answer but complete steps: booking a placement test, collecting a document, flagging a missing fee.

  • Caution on hype. Gartner also predicts that over 40% of agentic AI projects will be canceled by the end of 2027, due to cost, unclear value or weak risk controls. The Georgia State lesson applies: start with narrow, measurable tasks.

  • WhatsApp as the front desk. Pounce worked because it used the channel students already used. In much of MENA, that channel is WhatsApp. Our view is that institutions answering in Arabic and English on WhatsApp will have the same advantage Pounce had with text messages.

How does Adawaty help?

We build and run the admissions assistant for you, using the same design principles the research supports:

  • AI Admissions Desk: an assistant on web, WhatsApp and phone, in Arabic and English. Every enquiry is logged as a lead in your Adawaty Campus pipeline, with a weekly report on questions answered, leads captured and gaps. Live in 10 days, $990 + $149/mo.

  • Campus: our student information system, so the assistant can work from real student status, as Pounce did.

  • Paperless Operations Sprint: if your enrollment steps still run on paper and spreadsheets, we automate 3 to 5 workflows end to end. $1,890 + $190/mo.

  • Guarantee: on time, or it's on us.

We won't promise you Georgia State's numbers. They came from one university's trial. What we can promise is the same structure: personal, timely, data-linked outreach with a human behind it.

See every package on our services page.

FAQ

What is summer melt?

Summer melt is when admitted students who intend to enroll don't show up on the first day. Researchers estimate it affects 10–20% of college-intending students each year, with higher rates among low-income and first-generation students.

Did the Georgia State chatbot replace admissions staff?

No. It handled routine questions and reminders. Questions it couldn't answer went to staff, and only 13.5% of students in the trial sent one of those.

Does an AI admissions assistant work for schools, not just universities?

The Georgia State evidence comes from a university. The principles, fast answers, reminders tied to real status and a human handoff, apply to any institution with an admissions process. Our AI Admissions Desk is built for schools, institutes and universities.

How much does Adawaty's AI Admissions Desk cost?

$990 setup plus $149 per month for care. It's live in 10 days.

Related notes

Sources

  • Page, L. C., & Gehlbach, H. (2017). "How an Artificially Intelligent Virtual Assistant Helps Students Navigate the Road to College." AERA Open. journals.sagepub.com

  • Gehlbach, H., & Page, L. C. (2018). "Freezing 'summer melt' in its tracks: Increasing college enrollment with AI." Brookings. brookings.edu

  • Page, L. C., Meyer, K., Lee, J., & Gehlbach, H. (2023). "Conditions under which college students can be responsive to nudging." EdWorkingPaper 20-242. edworkingpapers.com

  • Meyer, K., Page, L. C., et al. "Let's Chat: Leveraging Chatbot Outreach for Improved Course Performance." EdWorkingPaper 22-564. edworkingpapers.com (PDF)

  • Georgia State University News, "Classroom Chatbot Improves Student Performance, Study Says," March 21, 2022. news.gsu.edu

  • EdScoop, "Higher education is warming up to AI, new survey shows" (2025 EDUCAUSE AI Landscape Study), February 17, 2025. edscoop.com

  • EDUCAUSE, "2025 EDUCAUSE AI Landscape Study." library.educause.edu

  • Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," June 25, 2025. gartner.com

Ready to stop losing admitted students?

Tell us your institution's name on WhatsApp and we'll map where applicants go quiet, and what to send them.

Start on WhatsApp