Watch Now
Finding the Right Moment: How UCF Reaches Students Before Problems Pile Up
How UCF Reaches Students Before Problems Pile Up
At the University of Central Florida, student success didn’t improve because the institution sent more messages.
It improved because UCF got far more precise about when to reach out, who to reach, and what kind of help actually mattered in that moment.
In a recent conversation, Tyler Walsh, Director of the Center for Higher Education Innovation (CHEI), shared how UCF shifted from reactive support to a signal-driven, upstream model—using conversational AI not as a broadcast channel, but as a scalpel.
What follows is not a story about technology adoption.
It’s a story about changing the order of work.
The problem: reactive systems create reactive work
Like many large institutions, UCF already had:
Dedicated advisors and support staff
Dashboards and reports
Emails, portals, and office hours
Yet help often arrived after damage was done:
A registration window missed
A hold unresolved until it blocked enrollment
A gateway course failed before anyone intervened
Support depended on students asking for help, or staff discovering issues once they had already escalated.
As Tyler put it, the challenge wasn’t effort. It was timing.
Tyler Walsh explains why UCF chose to develop Knightbot
The shift: from broadcast to precision
UCF’s breakthrough wasn’t replacing human work—it was reordering it.
Instead of asking:
How do we reach more students?
UCF began asking:
Which students are about to hit a barrier—and how much time do we have to intervene?
That reframing changed everything.
Conversational AI became useful not because it automated responses, but because it allowed UCF to:
Detect early signals of friction
Intervene before deadlines and consequences
Reserve human time for students who explicitly needed help
Email still goes to everyone.
Advisors still provide personalized guidance.
But text outreach is used surgically—for the students most likely to stall without timely support.
What “working upstream” looks like in practice
At UCF, working upstream means treating barriers as predictable events, not surprises.
Examples include:
Registration holds identified weeks before enrollment opens
FAFSA verification requirements known in advance
Course performance signals that appear well before failure
These aren’t guesses. They’re data-backed moments.
Instead of waiting for students to ask questions, UCF proactively reaches out with:
A relevant message
A clear call to action
An easy way to raise a hand for help
Tyler Walsh describes the real power of Knightbot’s proactive outreach to students
Signals, not assumptions
A critical insight from UCF’s work is that precision depends on signals, not segmentation stereotypes.
UCF doesn’t decide who needs help based on:
Demographics alone
Static risk labels
Broad outreach rules
Instead, it watches for real behavior and context:
Did the student respond to a message?
Did they indicate they want help?
Are they engaging (or not) when it matters?
Those signals shape:
Who gets contacted
When outreach happens
Which staff step in
This turns outreach into a conversation, not a guess.
Tyler Walsh discusses Knightbot’s ability to send targeted campaigns with relevant information to each learner.
Impact: fewer fires, better outcomes
One of the clearest examples Tyler shared involved registration holds.
By texting students before registration opened:
UCF resolved hundreds of holds in advance
More students registered on time
Engagement rates were significantly higher for students who interacted with the campaign
Students who responded and asked for help registered at substantially higher rates than peers facing the same barrier who did not engage.
The takeaway: early, relevant outreach doesn’t just inform, it changes behavior.
Staff experience: from cold calls to real conversations
The precision model changed staff work just as much as student outcomes.
Before:
Long call lists
Little context
Low response rates
After:
Lists of students who asked for help
Clear understanding of the issue before outreach
Students who expect (and answer) the call
As Tyler described, this isn’t just efficiency. It’s morale.
Tyler Walsh describes the impact of Knightbot’s outreach on staff morale
Ethics and trust: making scale safe
UCF’s approach only works because trust is protected.
From the start, the institution built guardrails:
Opt-out messaging to respect student choice
Institution-owned, curated knowledge—not free-form answers
Clear escalation paths for safety, mental health, housing, and food insecurity
Careful, limited piloting of generative AI
Trust isn’t a side benefit—it’s infrastructure.
Tyler Walsh shares the importance of ensuring Knightbot’s messages build trust
The deeper lesson for institutions
UCF’s story isn’t about copying a tool or a campaign.
It’s about:
Re-centering work around moments that matter
Letting data and conversation signals guide effort
Using AI to surface human work, not replace it
Designing for your institution’s scale, staffing, and students
As Tyler emphasized, this approach looks different everywhere, and it should.
According to Tyler, the first step leaders should take when considering a similar approach is identifying the problem or issue unique to their own institution.
UCF didn’t win by doing more.
They won by doing earlier, smarter, and more human work, enabled by precision tools that respect trust, staff capacity, and student reality.
That’s what finding the right moment really means.