Proactive pre-enrollment support
What if every student who said yes had help making it all the way to day one?
At Georgia State University, researchers tested an AI-supported, two-way texting intervention designed around the actual steps students needed to complete between commitment and enrollment.
Among students who had already committed to Georgia State, those assigned to Pounce were 3.3 percentage points more likely to enroll on time—a 21% reduction in summer melt.
At the scale of an incoming class of approximately 3,500 students, the study authors estimate that effect would mean about 116 additional students matriculating.
The intervention was delivered through Pounce, developed by AdmitHub, now Mainstay.
A new conversation begins
The result behind the result
High response was not enough.
Students used Pounce whether or not they had already committed to Georgia State. But the outcomes were very different.
Still deciding
83.9%
responded at least once. These students had been admitted, but had not yet told Georgia State they intended to enroll.
Treatment-group n = 2,770.
Already committed
88.5%
responded at least once. These students had already said yes to Georgia State and still had steps to complete before arrival.
Treatment-group n = 975; 1,948 committed students were randomized overall.
Where should I enroll?
What do I need to do next to make it to day one?
Both groups talked to Pounce. The meaningful outcome appeared once the intervention matched the job the student was trying to do. For a committed student, that job had become concrete—and often visible in institutional data.
Designed intervention
The intervention started before the reminder.
Pounce was introduced early as part of the student’s Georgia State experience—an official guide students could recognize, respond to, and return to throughout the transition.
From there, current student context could shape a relevant conversation, a clear next action, an immediate answer, or a handoff to a person.
- Expected
- The learner knows who is reaching them and why.
- Relevant
- The conversation reflects something true right now.
- Actionable
- The learner can understand what to do next.
- Conversational
- The learner can reply in the same channel.
Establish the relationship
Introduce who is speaking, why the channel exists, and how the student can use it for help throughout the transition.
Know where the student is
Committed, undecided, or going elsewhere are different student states with different jobs to be done.
Know what is unfinished
Use regularly updated information about actual progress rather than treating every admitted student the same.
Reach out about a real barrier
Center support on something true for that learner: a transcript, aid requirement, counseling step, hold, or another consequential task.
Make the next action clear
Help the student understand and complete the next useful step, rather than merely announcing that a requirement exists.
Make it easy to answer
Let students reply without navigating another portal, downloading an app, creating a login, or knowing which office to ask.
Answer what can be answered. Escalate what cannot.
Pounce handled common questions and routed unanswered questions to Georgia State staff through the same conversation.
Measure whether behavior changed
Response matters. So do completed steps, resolved holds, enrollment, and the measures that do not move.
This is the difference between sending messages and designing an intervention.
From signal to conversation
See what Georgia State actually did.
Between April and August 2016, Georgia State and AdmitHub used branching message flows across more than 90 enrollment topics. Students did not need all 90. Current student-level information determined which support was relevant.
April · newly admitted
Establish an official channel
Pounce introduced itself as the student’s official Georgia State guide—a place to ask questions, get help staying on track, and continue the conversation throughout the transition.
- Student state
- Newly admitted to Georgia State
- Signal
- Admission and permission to receive text messages
- Why it matters
- The first message established who was speaking, why the student was hearing from them, and what the channel was for.
- Evidence
- The study reproduces this historical opening message. It did not independently isolate the welcome as a treatment component.
The opening message is verbatim from the paper; the reply and continuation are illustrative.
From action to outcome
The gains appeared before enrollment did.
Enrollment was the final outcome. Before that, the full Pounce intervention moved several of the steps that make enrollment possible among students already committed to Georgia State.
Final transcript
+4.3 pp
more submitted a final high-school transcript
Immunization
−4.8 pp
fewer still had an immunization hold
Loan acceptance
+5.5 pp
more accepted a student loan
Loan counseling
+6.3 pp
more completed loan counseling
Stafford loan
+6.9 pp
more accepted a Stafford loan
On-time GSU enrollment
A conversation students used
Students answered. Human attention stayed focused.
Among committed treatment students, Pounce became an ongoing source of support—without turning every exchange into equivalent staff work.
The goal is not to maximize message volume. It is to make useful support available at the moment a student needs it, while preserving human judgment for the situations that require it.
What the research taught us
We start with the moment, not the message.
Georgia State helped establish something that has continued to shape Mainstay’s approach: a conversation can be easy to start and still fail to change an outcome.
Who is reaching the learner? What is true right now? What action can they take next? What happens when they reply? How will we know whether anything changed?
Establish the relationship.
Make the channel recognizable, legitimate, and useful before asking it to carry consequential interventions.
Start with the outcome.
Define what should change before deciding what to send.
Define the student state.
Know whose problem the intervention is trying to solve.
Use a real signal.
Let current context determine when support is relevant.
Make the next action clear.
Give the learner something concrete they can do.
Keep conversation easy.
Reach people in a channel that makes responding natural and requires no separate app or login.
Preserve human judgment.
Use AI where it creates capacity. Keep people close to the moments that require them.
Learn from the result.
Measure behavior, outcomes, and the places where the intervention did not work as expected.
Mainstay’s approach has continued to develop through additional research, research and institutional partners, behavioral science, practitioners, our community, conversational-design expertise, years of learner interactions, and modern conversational AI and escalation capabilities.
We have the research. We have the implementation experience. And we’ll share what we’ve learned.
A practical starting point
You don’t need 90 workflows to start.
Georgia State ultimately built support across more than 90 enrollment topics. Your first intervention can be much smaller.
Start with one consequential moment where the learner is known, the barrier is visible, the next action is clear, and timely support could still change what happens.
- The population is clearly defined.
- The learner recognizes the communication channel.
- The institution can see the student’s status.
- An unfinished requirement matters.
- The next action is knowable and measurable.
Start with the barrier
Where are committed students getting stuck?
Maybe students are waiting on financial-aid verification. Maybe transcripts, orientation, or health requirements are slowing progress.
Maybe staff can already see where students are stuck but cannot personally reach everyone at the moment it matters.
Let’s start there.