Walk into your internship already
knowing how to do the job.
In week one you pick a real company — a public company you'll research deeply, or a live local business as your client. Everything you build for eight weeks compounds on it.
By Demo Day you don't have eight exercises. You have one coherent engagement: an intelligence system, verified analysis, customer insight, a campaign, an automation with production data — and a recommendation you can defend under questioning.
AI use is unlimited — and fully disclosed. The graded skill is direction and verification, not typing.
20% instruction, 70% building, 10% presenting. Every session opens with a business problem, never a concept.
Every number traced to a source. Every claim cited. Every limitation disclosed. Like the job.
Six deliverables that stack.
One portfolio, one real business — each deliverable builds on the last, which is why Demo Day feels like one engagement instead of eight assignments.
Research Assistant + Engagement Charter
Before the deliverables: build your Market Research Assistant, learn why a blank chatbot fails, and pick the real company your whole engagement will run on. Everything below stacks on this choice.
Company Intelligence Copilot
An AI system grounded in your company's real documents — filings, transcripts, reviews — that answers questions with citations, and says so when the answer isn't in the sources.
Verified Financial Analysis Memo
AI reads the statements; you check the math. Two pages, a ratio panel, three flagged risks — every figure footnoted to its source, with a verification log.
Customer Insight Dashboard + Data Story
Hundreds of reviews and survey responses turned into an honest dashboard and an 800-word insight narrative — including one finding that contradicts the obvious take.
Brand-Voice System + Campaign Kit
Not ten generic AI posts — a reusable voice system derived from the company's real materials, executing a full campaign: ads, social, landing page, email, measurement plan.
Business Automation, in Production
A guarded automation that runs live for the final two weeks of the course. You present measured impact data — time saved, with the run log to prove it.
The Recommendation, Defended
A two-page recommendation memo answering your engagement's business questions — presented at a public Demo Day and defended in live Q&A before a panel that includes a business owner and someone who hires.
Eight weeks, week by week.
One 2.5-hour session per week plus ~2 hours of independent build time. Click any week.
The Analyst's Copilot
"You have a case interview tomorrow: should Chipotle launch breakfast?" Learn why a blank chatbot fails, then build a reusable Market Research Assistant encoding real frameworks — and pick your focus company.
Company & Competitive IntelligenceGraded audit №1
Ground AI in your company's real filings and transcripts. Then the first graded audit: classify ten AI "findings" as supported, distorted, or fabricated — with the source passage as proof. Build a cited competitive positioning table.
Financial Analysis, VerifiedGraded audit №2
AI narrates the 10-K; your spreadsheet checks it. Find five seeded errors in an AI-written financial summary — wrong growth rate, mixed fiscal years, invented segment figure. Then write your own memo where every number traces.
Customer & Market InsightGraded audit №3
"The owner says people love us but complain about price. Is that what the data says?" Theme-code real review data with AI, hand-audit its coding, then catch a peer's AI-assisted analysis overclaiming — in writing.
Marketing & Brand Systems
Why ten AI posts in ten minutes are generic garbage — and what a professional builds instead: a brand-voice system derived from the company's real materials, executing a full campaign with a measurement plan that says what would kill it.
Business Automation
Audit the workflows, pick the target, and launch a guarded automation — live before you leave the room, because it needs two full weeks of production runtime before Demo Day. Draft-don't-send is a hard rule for anything customer-facing.
Assemble & Harden
Everything becomes one engagement package, led by a two-page recommendation memo. Then real user testing — your client, or a peer playing the skeptical executive: "show me where that number comes from." Fix, guard, or disclose.
Demo Day
Seven minutes, then three of panel Q&A — with your copilot answering a panel question live and your automation's real production data on screen. The panel includes a business owner, faculty, and someone who hires.
You are graded on catching AI's mistakes —
because that's the skill employers can't find.
Three formal audits, spaced across the course. Each one plants real AI failures in front of you and scores whether you find them.
The Citations
Ten AI-generated findings about a real document set. Which are supported, which distorted, which entirely fabricated? Prove it with the source passage — or its absence.
The Math
An AI-written financial summary of a real filing with five seeded errors. Find the wrong growth rate, the mixed fiscal years, the invented figure — and correct them against the filing.
The Analysis
Swap datasets with a peer, run the same analysis, and identify where their AI-assisted conclusions overclaim what the data can support. Adjudicated by the instructor.
Try it now: one of these findings is fabricated.
An AI summarized the annual report of Cascade Coffee Co. (a practice company from the course). Two findings below are supported by the report. One is entirely invented. Click the fake.
Résumé bullets with receipts.
Selected Experience
- Built a document-grounded AI research copilot for [company], answering diligence questions with citations across 10-K filings and earnings transcripts.
- Produced a verified financial analysis — ratio panel and risk flags with every figure traced to source — and a customer-insight dashboard from 300+ reviews.
- Shipped a guarded LLM automation that ran in production for 2 weeks, saving [X] hrs/week — impact data logged and presented to a business panel.
- Delivered and defended a growth recommendation before a panel including the business's owner.
A case-interview story that's true
The Demo Day memo is a ready-made "walk me through an analysis you've done" answer — with exhibits.
Claims you can defend live
Panelists ask "where does that number come from?" — and tracing it on the spot is part of your grade. Interviews feel familiar after that.
Every concentration
The engagement flexes to your focus:
If you can build a slide deck, you can take this course
No coding, no CS prerequisite. An optional Builder Track rebuilds your copilot and automation in Python for those who want the technical depth.
One curriculum, four formats.
| Format | Duration | Scope |
|---|---|---|
| Weekend Intensive | 2 days | Research copilot + competitive intelligence + a compressed campaign sprint |
| Analyst Core | 4 weeks | Weeks 1–4: copilot, verified financials, insight dashboard |
| Full Engagement | 8 weeks | The complete arc as described, with production automation and Demo Day |
| Semester Elective | 14–16 weeks | Full engagement + a second client cycle and deeper Builder Track |
Frequently asked
No. The default path is entirely no-code — assistants, projects, spreadsheets, Looker Studio, and Zapier free tiers. The optional Builder Track rebuilds your copilot and automation in Python for students who want it, framed as "rebuild it properly."
Our rule is the inverse of most classrooms: AI use is unlimited and fully disclosed. What's graded is direction and verification — three of your grades come specifically from catching AI being wrong. An undisclosed fabricated figure is an integrity failure; a disclosed limitation is professionalism.
Any public company with deep documentation works — you'll research it like a diligence analyst. Where possible we pair students with live local or campus businesses, whose owners join Demo Day as your client.
Analytics is one week of this course. The engagement spans research, financial analysis, customer insight, marketing systems, and automation — the actual breadth of a junior business role — with AI as the through-line and verification as the standard.
Founding-cohort pricing and the first calendar are being finalized now. Requesting a seat puts you first in line and includes an invite to a free AI Career Copilot mini-workshop.