How Colaberry Built an AI Mentor to Support 1,000+ Students in Real Time

Every student deserves a mentor. But in most online learning systems, mentorship is the first thing to disappear.
 
Discussion boards go silent. LMS reminders go unread. Students drop off not because they lack ability, but because no one noticed they were falling behind.
 
At Colaberry, we faced the same challenge supporting over 1,000 active learners. Human mentors, no matter how dedicated, can’t scale fast enough to check in with every student exactly when they need support.
 
That’s why we built an AI mentor agent—designed to monitor key indicators, send timely personalized messages, and follow up in a way that feels human, helpful, and just in time.
 
This is the story of how we did it.

Designing the System

When we started building the AI mentor, our first principle was this: scale should never come at the cost of empathy.
 
We didn’t want automation that spammed students with robotic messages. We wanted an agent that felt personal, helpful, and responsive—no matter how many students it was supporting.
 
The system we built works in three core layers:

1. Trigger Logic: SQL-based rules monitor real-time student behaviour (like logins, attendance, or progress).
2. Personalized Prompts: If a rule is tripped, the agent pulls a tailored prompt with student-specific variables (e.g., name, module, pace).
3. Goal-Driven Messaging: The AI reaches out via the best-fit channel (WhatsApp, email, or in-app chat) and follows a branching conversation designed to help the student get back on track.

The Components of the AI Mentor

We didn’t build one big AI—we built a system of small, intelligent parts that work together.

Here’s what makes it run:

1. SQL Trigger Engine
Every student action (or inaction) is logged in a central database. We created dynamic SQL rules to detect when a student’s behaviour signals a need—for example:
 

• Not logging in for 5 days
• Below 70% pace
• Missed a checkpoint milestone

2. Prompt Personalization System
Once a rule is tripped, the system fetches a personalized message template from our prompt library. It fills in the student’s name, recent activity, and a tone that matches the moment (encouraging, supportive, or action-focused).


3. AI Agent Communication Layer
This layer sends the message via the most effective channel:
 

• WhatsApp
• Email
• In-app Assistant

Then, it manages a real-time dialogue, escalating if needed or closing the loop if the student takes action.
 

4. Feedback Loop + Logging
Every interaction is scored by outcome:
 

• Did the student reply?
• Did they re-engage with the platform?
• Did their KPI improve?

This data improves future prompts and helps us tune the system for greater impact.

 

What It Feels Like for the Student
 
From the student’s perspective, it doesn’t feel like a system.
It feels like someone actually noticed.
 
Instead of another generic LMS nudge—“You have 3 assignments overdue”—they get a message that says:
 
“Hey Alex, I noticed you haven’t logged in this week. Want help getting back on track?”
 
If they respond, the conversation continues:
 

• The agent might offer encouragement: “You’ve made great progress so far. Let’s finish strong.”
• Or it might provide logistics: “Click here to schedule a 1:1 support session.”
• Or even accountability: “Would it help to set a small goal for today?”

The tone is warm. The timing is relevant. And most importantly, the student feels seen.
 
This small shift—from automation to meaningful engagement—changes everything.

 

Results at Scale
 
Once we deployed the AI mentor to our full cohort of 1,000 students, the results were immediate.
 
In just the first month, we saw a measurable shift in behavior:
 

• 28% increase in weekly logins
• 22% boost in students staying “on pace” with assignments
• 39% response rate to AI-initiated messages

More importantly, students began re-engaging because the system felt human—not like a dashboard nagging them, but like someone who actually cared.

And from a support team perspective, the AI mentor handled over 3,000 personalized interactions without requiring a single added human hire.

It was proof that empathy can scale—if you build the system right.

 

 

What We Learned

Building an AI mentor at scale wasn’t just a technical project, it was an emotional design challenge.

Here are 3 key lessons we took away:

1. Timing is Everything
Messages sent too late feel irrelevant. Messages sent too often feel annoying. The sweet spot triggered outreach that feels like it noticed you at just the right moment.


2. Tone > Technology
Students don’t care that it’s AI. They care that it sounds like someone who understands them. We found that a warm, human tone increased reply rates by over 30%.

3. Feedback Loops Create Intelligence
The AI mentor improves over time because every message, every click, every outcome feeds the system. That’s how it grows smarter—and more helpful—with every student interaction.


This isn't just a tool—it’s a support layer. One that never sleeps, never judges, and always nudges toward success.

 

What’s Next
 
This first deployment was just the beginning.
 
We’ve already started expanding the AI mentor’s capabilities across new domains:
 

• Career Support: Helping students build resumes, prep for interviews, and track job applications
• Behavior Nudges: Reinforcing healthy learning habits with gentle accountability
• Custom LMS Integrations: Embedding our system into platforms used by bootcamps, universities, and workforce boards

We're also exploring ways to open up the system to others:
 

• An API-based integration layer for partners
• A prompt marketplace for curated conversation logic
• A dashboard for live feedback and tuning

The dream? That one day, every student—not just the lucky few—has someone (or something) looking out for their success.
 
And we’re building it.
 

Let’s Build This Together

If you’re running an LMS, bootcamp, or online education program, you’ve probably asked:

“How can we support every learner without hiring 10x the staff?”

This blog was the answer to that question.

We’ve already built the system. We’ve tested it with 1,000+ learners. And we’ve proven it works.

Now we want to bring it to your organization.


Want to see how it could work for you?

 

Let’s talk.

We’ll walk you through the architecture, show you real student examples, and explore how to launch a pilot inside your platform.