brianletort.ai
← Library

Georgia State University

Mixed evidenceExceptions onlyQueue elimination

Pounce summer-melt support

Enrollment support does not require counselors to initiate and answer every routine student interaction; a task-aware assistant can provide personalized outreach while routing exceptions to staff.

Enrollment · Atlanta, Georgia, United States

Collections: Human still decides · Queue eliminated

An editorial scene for Georgia State University contrasts track multiple enrollment requirements and seek help when a task or deadline is unclear. with use university records to send task-specific deadline reminders and offer help. in the pounce summer-melt support workflow.

Executive brief

The operating-model shift, in one view.

The strongest evidence comes from changing the handoff, not merely adding a chatbot: routine task guidance moved to the assistant, while counselors became the exception path.

AI value · On-time fall enrollment among GSU-committed students

Verified

Students assigned to Pounce were 3.3 percentage points more likely to enroll on time, equivalent to a 21% reduction in summer melt in the peer-reviewed study.

The 3.3-point effect applies to the GSU-committed subgroup, not all accepted students; the primary GSU page rounds the relative melt reduction to 22%.

Before

Track multiple enrollment requirements and seek help when a task or deadline is unclear. → Answer routine and complex questions and provide individual summer outreach.

After

Use university records to send task-specific deadline reminders and offer help. → Answer common student questions automatically. → Handle questions that require professional judgment or information the assistant lacks.

Human boundary

The assistant provides routine guidance; counselors and advisors retain authority for unresolved, exceptional, or consequential student issues.

Why it matters

Enrollment support does not require counselors to initiate and answer every routine student interaction.

How the work changed

Before

How the work ran before the change.

  1. Step 1 of 2

    Prospective student

    Track multiple enrollment requirements and seek help when a task or deadline is unclear.

    ControlStudent initiates contact under the university's standard process.

  2. Step 2 of 2

    Admissions counselor

    Answer routine and complex questions and provide individual summer outreach.

    ControlCounselor determines the guidance.

What changed

Enrollment support does not require counselors to initiate and answer every routine student interaction.

Decision rightAI handles the default; humans own exceptions

After

How the same work runs now.

  1. Step 1 of 3

    Pounce

    Use university records to send task-specific deadline reminders and offer help.

  2. Step 2 of 3

    Pounce

    Answer common student questions automatically.

  3. Step 3 of 3

    Admissions counselor

    Handle questions that require professional judgment or information the assistant lacks.

    ControlCounselors retain exception handling and authoritative guidance.

Process model built from the published workflow evidence for Georgia State University. Every step, actor, and control appears in full below.
Every step, actor, and control

Exception path

Questions the assistant cannot answer are routed to Georgia State counselors and advisors.

Work removed

  • Routine deadline reminders sent individually by counselors
  • Counselor responses to the majority of common enrollment questions
  • A material share of repetitive pre-matriculation messaging

Decision authority

The assistant provides routine guidance; counselors and advisors retain authority for unresolved, exceptional, or consequential student issues.

Before

  1. 01

    Prospective student

    Track multiple enrollment requirements and seek help when a task or deadline is unclear.

    Control: Student initiates contact under the university's standard process.

  2. 02

    Admissions counselor

    Answer routine and complex questions and provide individual summer outreach.

    Control: Counselor determines the guidance.

After

  1. 01

    Pounce

    Use university records to send task-specific deadline reminders and offer help.

    Control: Outreach is tied to each student's incomplete enrollment tasks.

  2. 02

    Pounce

    Answer common student questions automatically.

    Control: Only a small fraction of questions are routed to staff.

  3. 03

    Admissions counselor

    Handle questions that require professional judgment or information the assistant lacks.

    Control: Counselors retain exception handling and authoritative guidance.

Work that left the path

  • Routine deadline reminders sent individually by counselors
  • Counselor responses to the majority of common enrollment questions
  • A material share of repetitive pre-matriculation messaging

Human role before

Counselors provided individualized reminders and answered both routine and complex enrollment questions, limiting practical summer caseloads.

Human role after

Counselors focus on the smaller set of questions that require professional guidance while the assistant handles routine outreach and answers.

AI roleTrigger personalized enrollment-task outreach, answer common questions, and route unresolved questions to university staff.

Outcomes

On-time fall enrollment among GSU-committed students

Verified

GSU's business-as-usual control group.→Students assigned to Pounce were 3.3 percentage points more likely to enroll on time, equivalent to a 21% reduction in summer melt in the peer-reviewed study.

April through August 2016. · 7,489 accepted students with eligible mobile numbers and text consent; 3,745 assigned to Pounce and 3,744 to usual process, with the reported enrollment effect in the pre-stratified GSU-committed subgroup.

The 3.3-point effect applies to the GSU-committed subgroup, not all accepted students; the primary GSU page rounds the relative melt reduction to 22%.

What leaders can reuse

Anti-pattern

Using a general chatbot without task-state integration, an explicit staff escalation path, or subgroup-aware outcome measurement.

Questions

  1. 01Which routine questions can be automated safely?
  2. 02What student signals should trigger human outreach?
  3. 03Will the evaluation distinguish intended users from people unlikely to use the service?

Portability conditions

  • Reliable task-level student data
  • Clear boundaries for routine versus counselor-required guidance
  • Consent and messaging controls
  • Randomized or otherwise credible outcome measurement

Reputation risk

low: label the 2016-era conversational AI accurately and do not imply modern generative-AI capabilities.

Evidence and authority

What the public record supports.

Current · updated

1 independent, 1 peer reviewed, 2 primary; publication outcomes are verified and reported.

Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 5b2556715b7a65ac

Related transformations

More in Enrollment →

New York City Fire Department · Exceptions only

FireCast risk-ranked inspection lists through RBIS

Fire companies can prioritize inspections by predicted risk instead of fixed cycles and local knowledge.

Waymo · Exceptions only

Fully autonomous Waymo One driving

Within a defined domain, automated driving can control the vehicle while remote assistance handles exceptions.

Ocado Group / Ocado Retail · Exceptions only

On-Grid Robotic Pick at Luton CFC

Robotic arms can pick more grocery items on-grid while people handle exceptions and fulfillment operations.

JPMorgan Chase · Exceptions only

COiN agreement extraction

Defined attributes can be extracted across annual volume while lawyers handle interpretation exceptions.

Sources

Read the evidence, freshness, caveat, and version policy.