MSBAi Learning Infrastructure — Pilot Brief

For the 8-person pilot team. April 2026.

The Big Idea

Most teams use AI to get things done faster. We’re using it to get smarter.

The MSBAi K-ai experiment is two things at once:

  1. A program coordination tool — K-ai manages the knowledge base, tracks decisions, answers stakeholder questions
  2. A live learning system — every interaction with K-ai is data about the program, the team, and human-AI collaboration

This brief describes how we’re building the infrastructure to capture that learning — and why it matters beyond MSBAi.

Two Frameworks Shaping Our Approach

1. Philosophy Eats AI

Schrage & Kiron, MIT Sloan Management Review, January 2025

The central argument: AI value creation depends less on technical capabilities than on philosophical clarity — specifically three dimensions that most organizations ignore:

In 2024, Google’s Gemini image model generated historically inaccurate outputs — depicting the US Founding Fathers as racially diverse, Vikings as Asian women — because its diversity training objectives overrode its accuracy objectives. Google never reconciled the conflict. The result wasn’t a technical bug — it was a teleological failure: a powerful system producing confident wrong answers because no one had clearly defined what it was for.

“Pattern matchers without purpose, computers that generate the wrong answers faster.”

MSBAi implication: K-ai is useful only if we’re clear about what we’ve trained it to do and why. The outer loop is how we keep that clarity — and sharpen it over time.

2. How to Reap Compound Benefits from Generative AI

MIT Sloan Management Executive Education, April 2026

The central argument: Organizations that treat AI as a capability accelerator — not just a productivity tool — outperform by building systematic learning on top of AI usage. The mechanism is a three-step cycle:

Verification → “Does this output meet the standard?” Catches errors. Necessary but insufficient alone. Produces binary answers.

Evaluation → “What does this output reveal?” Requires domain expertise. Surfaces tacit knowledge. Discovers new standards. This is where competitive advantage is built.

Learning Capture → “How does this insight persist?” Converts single insights into organizational knowledge — updated prompts, documented criteria, shared repositories. Version control for organizational judgment.

“Organizations combining strong organizational learning with AI-specific learning are up to 80% more effective at managing uncertainty. Yet as of 2024, while 70% of companies adopted AI, only 15% used it for organizational learning.”

MSBAi implication: We’re in that 15%. The pilot group’s job is to run the outer loop — not just use K-ai, but evaluate what K-ai reveals and capture what we learn.

The Three-Level Structure

┌──────────────────────────────────────────────────────┐
│  EXTERNAL OBSERVER                                   │
│  Prof. Willie Ocasio (Towey Professor, Gies)         │
│  Applies organizational theory to interaction        │
│  patterns. Weekly digest. Teaches by questioning.    │
└──────────────────────────────────────────────────────┘
                          ↑
┌──────────────────────────────────────────────────────┐
│  OUTER LOOP — The Learning Layer                     │
│  8-person pilot team + Amber Glynn + Vishal Sachdev  │
│  Verify → Evaluate → Capture                         │
│  Weekly cadence                                      │
└──────────────────────────────────────────────────────┘
                          ↑
┌──────────────────────────────────────────────────────┐
│  INNER LOOP — K-ai Operations                        │
│  Receives inputs → Updates KB → Responds → Logs      │
│  Real-time, every interaction                        │
└──────────────────────────────────────────────────────┘

The Inner Loop (K-ai — already running)

K-ai operates across three channels:

Every interaction is logged to the audit trail in the program’s GitHub repository. Every decision, action item, and open question is extracted and filed. Every commit is permanent and auditable.

What K-ai manages: program decisions, course design, faculty coordination, stakeholder Q&A, curriculum knowledge.

What K-ai doesn’t do: make strategic calls, resolve ambiguous decisions, override Vishal.

The Outer Loop — Your Role

What the pilot team does

After interacting with K-ai, send a short observation to msbai@illinihunt.org. Free text — no template required. The signal matters more than the format.

Three useful questions to answer (any one is enough):

That’s the verification layer. K-ai classifies and files your observation automatically.

What happens with your observations

Amber Glynn reviews the operational patterns weekly:

Vishal Sachdev reviews the strategic and content patterns weekly:

Vishal owns all learning capture — updates to K-ai’s instructions, lessons learned, KB improvements. One person commits; the system stays coherent.

Cadence

What Who When
Use K-ai, send observations 8-person team Async, as interactions happen
KB changes digest (what was updated this week) K-ai → Vishal Daily 8am
Pilot interaction summary (non-Vishal interactions) K-ai → Vishal Daily 8pm
Operational review Amber Weekly
Strategic review + commits Vishal Weekly
Interaction pattern digest K-ai → Ocasio Every Friday 9am

Willie Ocasio — External Observer and Research Partner

What happened

On the night of April 17-18, Professor Willie Ocasio (James F. Towey Professor, organizational theory) encountered K-ai for the first time and sent ten questions over 53 minutes — from “explain the AI process” to “I want to use this as a research site.”

In that single conversation, he independently:

The full exchange: https://msba-online.pages.dev/docs/ocasio-ai-governance

His role going forward

Professor Ocasio is the external observer layer of the learning infrastructure. He is not running operations. He is applying organizational theory to what the system produces.

His specific value: The pilot team evaluates operational quality — did K-ai handle this correctly? Professor Ocasio evaluates structural and theoretical implications — what does the pattern of interactions reveal about how AI reshapes organizational attention, knowledge flows, and authority?

Mechanism: K-ai sends him a weekly digest (Fridays) of interaction patterns, designed to surface interesting observations and invite his questioning. He replies by email. His questions become evaluation data — what he probes reveals what the system needs to explain better. He teaches K-ai by asking good questions.

Research trajectory: Over 12-18 months, the interaction data becomes an empirical dataset for research on AI-mediated organizational coordination. MSBAi is both the subject and the collaborator.

Note: Professor Ocasio’s formal invitation to this role is pending — Vishal is reaching out directly.

Why This Matters Beyond Program Coordination

MSBAi is building a graduate program about AI-augmented analytics. We are simultaneously using AI to build the program. That’s not a coincidence — it’s the point.

Every insight the outer loop produces about human-AI collaboration is direct curriculum content:

The pilot group isn’t just testing a tool. You’re co-developing the evidence base for a curriculum about AI in organizations.

How to Start

  1. Use K-ai — email questions, send updates, ask about the program. Interact naturally.
  2. Send one observation after each interaction — what worked, failed, or was interestingly wrong. Email to msbai@illinihunt.org.
  3. Telegram pilot group — tag @MSBAiBot in your message to get a response. The group is shared so everyone can see the exchange.
  4. Expect a Friday digest from K-ai to Professor Ocasio — you’ll see his replies come back into the system as his research observations build.

No meetings required. No templates to fill out. The bar is low: one email observation per interaction is enough to generate signal.

Reference Materials

K-ai is the AI coordinator for the MSBAi program at Gies College of Business, UIUC. Questions: vishal@illinois.edu