Student AI Companion

Every MSBAi student gets a personal AI companion that knows their entire program — all courses, all due dates, all connections between what they’re learning simultaneously. It lives on their phone. It reaches out before they think to ask.

Status: Proof of concept running now → operationalize Summer 2026 → live at Fall 2026 launch

The Problem It Solves

Online graduate students experience programs as a series of isolated courses. They forget what they learned two weeks ago. They miss the connections between concurrent courses. They do exactly what’s assigned — and nothing more.

MSBAi is designed as a coherent 15-month program, not nine independent courses. The student AI companion makes that coherence tangible — surfacing cross-course connections, reinforcing learning between sessions, and nudging students toward the portfolio work that differentiates them.

“If all you’re doing is what we’re assigning you, you’re not doing enough.” — Vishal Sachdev

Faculty vs. Student Infrastructure

The same infrastructure that coordinates 20+ program stakeholders gets mirrored for students:

What the Companion Does

Pull: Students Ask, Companion Answers

One interface across all nine courses:

• “What do I have due this week?”
• “What did we cover in data modeling last week?”
• “When does FIN 550 start and what do I need to know going in?”
• “I’m stuck on the ETL milestone — what did we cover that’s relevant?”

The companion draws from a unified index of all course content, schedules, and announcements — something no individual instructor can provide alone.

Push: Intelligent Outbound Drip

The companion reaches out proactively, timed to the learning calendar:

Cross-course synthesis (what no Canvas notification can do):

“Last week in data modeling you learned normalization — why clean schemas matter by design. Ron’s storytelling course starts Monday. You’ll see exactly why that matters when you try to visualize messy data.”

Spaced reinforcement (combating the forgetting curve):

“Three weeks ago you built your first ETL pipeline. FIN 550 starts in two weeks. That same pipeline pattern is how you’ll pull Compustat data. Here’s a 5-minute refresh.”

Portfolio nudges (beyond the syllabus):

“You completed the Week 5 milestone. That’s the baseline every student hits. One more thing you could publish to GitHub this week would make your portfolio stand out to employers. Here’s what that looks like.”

Transition moments (course starts, summer break, final stretch):

“You’re starting your third semester Monday. Here’s what you’ve built over the last 10 months — and what the final 5 months will add to it.”

Spaced Repetition by Design

• Once-a-week courses create a forgetting gap between sessions
• Drip messages bridge the gap: recap → preview → one question
• Timed to the program calendar, not generic reminders

Proof of Concept: What’s Already Working

Running now in Vishal’s on-campus course:

The 11% active engagement rate is consistent with early-stage behavior change interventions. The remaining 89% receive reinforcement passively — the nudge reaches them even when they don’t respond.

Design Principles

Phone-first, not Canvas-first. Canvas announcements go unread. A WhatsApp or iMessage notification doesn’t. Opt-in is required; students choose their channel at enrollment.

Push + pull, both matter. Pull (student-initiated) gives them agency. Push (program-initiated) serves them before they know they need it. The combination is what makes it a companion, not just a search interface.

Cross-course synthesis is the differentiator. No individual instructor can see what a student is experiencing across three concurrent courses. The companion can. That synthesis — “here’s how this week’s data modeling connects to storytelling and algorithms simultaneously” — is the distinctive value.

Structured, not random. Every outbound message has a purpose tied to the learning calendar. Not motivational quotes. Not generic reminders. Contextual information that matters this week.

Alongside Canvas, not competing with it. Canvas remains the homepage for coursework. The companion is the layer that makes the program coherent between Canvas sessions.

Beyond compliance. The minimum is what gets assigned. The companion nudges students toward the above-minimum work — GitHub contributions, portfolio depth, cross-course synthesis — that employers actually distinguish.

What It Is Not

• Not a tutoring system. It knows what is being taught, not the domain content itself. For domain help, students have instructors and TAs.
• Not peer-to-peer. 1:1 bot-to-student only. Community and peer connection happen in Canvas, studio sessions, and cohort channels.
• Not mandatory. Opt-in. Students choose their channel or opt out.
• Not a Canvas replacement. It runs alongside Canvas, drawing from it via API.

Open Questions

• Can this interface serve dual purpose — knowledge acquisition and peer community? (Peer-to-peer would require different architecture, but the AI knowledge base could reduce load on community channels by answering factual questions.)
• Opt-in mechanics: how do students register their preferred channel at enrollment?
• FERPA considerations for Canvas API data use
• Who owns the outbound drip calendar — Vishal alone, or a shared instructional team workflow?
• Does the companion eventually surface inside Canvas as an embedded chatbot, or remain phone-only?

Timeline