Saved
↑↓ slide  ←→ focus
AI IN EDUCATION · APRIL 16, 2026

What's possible
in the classroom
today?

A working tour of what ships this semester and how it lands in class. Live demos, real outputs, and a clear map of what faculty can do before, during, and after class.
Presented by Mohammad Soltaniehha
Questrom School of Business Boston University
§ 00Opener · Icebreaker
01 / 27
THE TALK BEFORE THE TALK

Yes, another AI talk!

By now, you have sat through the pitch, the demo, and the 'this changes everything' moment. So have we. This one is different.

§ 01Why this matters now
02 / 27
THE PRODUCTIVITY PARADOX

Tools don't shift productivity.
Reorganization does.

Computers sat on every desk for a decade before they moved the productivity needle. What changed was not the hardware. It was how companies restructured around it.

§ 01Why this matters now
03 / 27
WHERE PEERS ARE NOW

The question has already shifted.

From "should we teach AI?" to "how do we make sure every student develops AI literacy?" From elective to foundational. From planning to implementation.

48
business schools
profiled by Inspire Higher Ed
47%
of business schools have
AI policies · AACSB
18mo
Leeds went from zero
to 100% core integration
9,000+
people trained
across Neoma
§ 05The 8 themes
04 / 27
Eight themes across 48 schools

How they're
organizing the work.

01
Infrastructure & ecosystems

Dedicated centers, secure AI platforms, and governance frameworks. Not ad hoc pilots but coordinated systems that scale.

02
Democratization

AI literacy moves from elective to required. Every student. Some schools extend to staff, alumni, and community.

03
Domain-specific AI

AI embedded inside marketing, finance, accounting, operations. Not a generic "AI course" but discipline-native applications.

04
Faculty development

Ongoing training, peer communities, microgrants, and safe spaces to experiment.

05
Responsible AI & ethics

Woven into every course, not siloed in one. Real cases with real consequences, not hypothetical scenarios.

06
Industry partnerships

Schools can't build everything alone. Partners provide tools, data, and real-world credibility.

07
Pedagogical evolution

AI tutors, orchestration platforms, personalized feedback at scale. Not just teaching about AI but teaching with AI.

08
Leadership

A dean or director owns the initiative. Top-down commitment creates urgency that bottom-up enthusiasm alone can't achieve.

§ 02What's newly possible
05 / 27
What's newly possible

Past the chatbot.

01

Agentic workflows

Give it a goal, it figures out the steps. A 2-week research task, done autonomously.

02

Tool integration

Works inside Gmail, Excel, and Word, not beside them. Your existing tools are AI-native.

03

Computer use

Sees and controls the screen like a person. Fills forms, navigates Blackboard, and books travel.

04

Deep research & reasoning

Multi-source, cited reports in 30 minutes. Thinks step-by-step through novel problems.

05

1M-token context

Ingests a full semester of readings at once. No more summarize-and-forget.

06

Multimodal & voice agents

Sees images, hears speech, talks back. Build a voice tutor or grade a whiteboard photo.

§ 03This is not hypothetical · Shipped last month
06 / 27
What shipped last week

What shipped,
not what's coming.

APR 11Anthropic

Claude, inside
Microsoft Word.

Native sidebar. Edits appear as tracked changes. In the doc faculty already write in.

PRODUCTIVITY→
APR 8SiliconANGLE

Anthropic launches
Managed Agents.

Agents become a first-class product. Development time: months to weeks.

AGENTS AS PRODUCT→
APR 9TechCrunch

The era of clicking buttons is over.

Bret Taylor, Sierra CEO: interfaces move from menus and nav to AI agents in plain language.

UI PARADIGM→
APR 13PwC

The AI gains are
concentrating. Fast.

A small group of firms is pulling sharply ahead. The rest are still experimenting. 1,217 executives · 25 sectors.

ECONOMIC EVIDENCE→
§ 04And in our own world · Higher ed last month
07 / 27
What higher ed did last week

Peers are already moving.
...Not all of it is working.

APR 11Harvard Crimson

HBS expands AI
across the MBA.

AI avatars, simulations, live exercises now in marketing, entrepreneurship, OB. Far beyond the required DSAIL course.

PEER PRECEDENT→
APR 6Inside Higher Ed

UChicago: $50M gift
for AI faculty.

Mansueto Faculty of Mind and Machine Challenge. Targets $200M, recruits 20 scholars across disciplines.

INSTITUTIONAL INVESTMENT→
APR 9Chalkbeat

Sal Khan: AI revolution "for a lot of students, a non-event."
"They just didn't use it much."

Three years after Khanmigo, most students ignore the tutor. Khan: the challenge is engagement, not capability.

REALITY CHECK→
APR 4CNN

Students are starting to sound the same.

Yale and others: homogenized classroom discourse. The sharp edge of overuse.

CAUTIONARY SIGNAL→
§ 4.5The work is already shifting
08 / 27
Coding after coders · NYT Daily

The work is
already shifting.

Clive Thompson interviewed roughly 75 developers. The majority now outsource most of their code to AI agents. At small startups, nearly everything new is AI-written. At Google, closer to half. A day's feature finishes in thirty minutes.

The caveat: mature codebases slow the curve. Google reports a more modest lift overall. Still, the day-to-day of programming has changed, and students are walking into that job market.

§ 06The debate nobody has resolved
09 / 27
The debate nobody has resolved

Two camps on de-skilling.

Camp A · The optimists

AI code is
actually more reliable.

Automates away rote human error. The AI doesn't forget the edge case. Students lifted to the architect role, not the typist.

"A student ships a working app in a weekend."

Upside: faster, cleaner, more ambitious work
Camp B · The worriers

Subtle bugs pile up
invisibly.

Students finish the assignment but lose the judgment that made it legible. The reckoning shows up as huge messes five years out.

"The same student can't debug it when it breaks."

Downside: long-term erosion of craft
This is the frame for every classroom decision. Short-term output lift versus long-term loss of judgment. Not ethics in the abstract, just that trade, repeated every time.
§ 07Key questions for Questrom faculty
10 / 27
Key questions

What happens
when the tools
get too good?

§ 08Student success · It already worked
11 / 27
Student success story
“
Claude served as an effective sounding board throughout our analysis. It helped us develop a strong starting point and better understand the analysis we wanted to perform. We now view Claude and GenAI as critical resources in both our academic and professional careers.
Wren, Nate, Kayla, Max · Questrom undergraduates · April 2026
§ 09The cheating reality · The flip side
12 / 27
Skip the obvious · Five methods you haven't heard about

Assessment
is broken.

1
Real-time overlay coaching
Cluely · FinalRoundAI

Listens on Zoom. Generates answers. Overlays them invisibly on the interviewer's screen. Founder suspended from Columbia; product now VC-backed at scale.

2
Smart-glasses exams
Ray-Ban Meta · Rokid

Photo the question. Cloud answers via earpiece. SAT banned smart glasses, March 2026.

3
AI "humanizers"
StealthWriter · Undetectable

Rewrite ChatGPT output with natural phrasing. Turnitin's AI score becomes gameable.

4
Autonomous course agents
Einstein · Feb 2026

Logged into Canvas, watched lectures, took quizzes, submitted essays. Instructure cease-and-desist.

5
LMS quiz-solvers & AI browsers
QuizSolve · Comet · Atlas

Extensions answer Canvas quizzes. Colorado State issued campus guidance.

§ 10Faculty research opportunity · Vibe Physics
13 / 27
Architect, not construction worker

Your skill
lies elsewhere.

"They are becoming less like construction workers and more like architects."

Clive Thompson · NYT Daily · Apr 14 2026
2 weeks
VS
a year's work

Harvard physicist Matthew Schwartz resumes the C-parameter Sudakov shoulder. A new factorization theorem. Published on arXiv, Jan 2026.

§ 11Everyday AI-adjacent productivity · Small wins that compound
14 / 27
Not the big model. The AI-adjacent tools I live in.

Five productivity tools I actually use.

Stt
Superwhisper
Dictate anywhere on the Mac. Email, notes, prompts. Huge time saver.
→
tts
Speechify
Turn articles and papers into audio. Catch up on reading while walking.
→
Link
Obsidian
Connects local markdown notes. Your data stays yours. A second brain, searchable forever.
→
Note
Zoom AI Companion
Clean summary after every meeting. Pipe it into Obsidian for a searchable archive.
→
THINK
Claude
Research, write, code, brainstorm, build. The one tool I never close.
→
§ 12Tool · Claude Cowork
15 / 27
Desktop AI · Anthropic Claude
Cowork.

A desktop AI that works inside a folder, produces real .docx, .xlsx, .pptx, .html, and more.

drag-and-drop files local-first multi-document context produces .docx / .xlsx / .pptx
§ 13Demo preview A · Polished documents
16 / 27
Case poster: Makatemia, AI and the Search for a Second Act
One case PDF in · a teaching package out

Every artifact,
same session.

DOCX
Teaching Note
DOCX
Discussion Questions
DOCX
Grading Rubric
DOCX
Supplementary Context
DOCX
Student Follow-Up
PPTX
Case Discussion Deck
XLSX
Case Data
XLSX
What-If Model
PNG
Case Poster (shown)
📖 case-teaching-cowork-playbook.md · the prompts behind these outputs click to open →
§ 14Demo preview B · Interactive classroom tools
17 / 27
Built with Cowork · Ships in a browser

Two apps. One class.

Board Plan

board-plan.html

Case Discussion Companion

case-discussion-companion.html
§ 15Tool · Claude Code
18 / 27
Terminal-first agent · Anthropic Claude
Code.

Where it really earns its keep for faculty: research. Data you never cleaned. Papers you never finished.

runs in your terminal reads and writes files runs commands the Vibe Physics engine
§ 16Tool · NotebookLM
19 / 27
Research-to-audio · Google NotebookLM.

Feed it papers. Get understanding back. Audio, summaries, study notes.

multi-source grounding audio overviews study guides always cited
§ 17Tool · Vibe Coding · Apps without code
20 / 27
Describe what you want · get a working app

Build without developers.

The emerging vibe coding category. Faculty can build custom classroom tools now. Sign-up sheets, cold-call randomizers, participation trackers, syllabus reviewers.

Lovable
Start here if you've never coded. Database and auth baked in. What we demo next.
lovable.dev · easiest
v0
A frontend generator by Vercel. The prettiest output. Pick it when the UI has to look polished for students.
v0.app · best design
Replit
The closest to real development. Pick it when you need backend logic, not just a pretty frontend.
replit.com · ships live
Also in the space: Bolt.new · Base44 · Claude Artifacts (zero setup)
§ 18Demo · Syllabus A-Eye · Built with Lovable
21 / 27
DEMO · VIBE CODING IN PRACTICE Syllabus
A-Eye.

Upload a syllabus. Get lecture-by-lecture feedback on where AI fits and where it breaks your assessment.

AI-proof this, AI-open that rewrite suggestions per-session commentary

"Horribly underserved by technology... now one person can say, I can build that for you."

Thompson on mid-sized firms · NYT Daily · Apr 14 2026
§ 19Showcase · CaseAlive.com · Mohammad's tool
22 / 27
5 minutes · my own tool · yours to use

Case PDF in.
Living conversation out.

Students chat with AI personas from the case. On-demand office hours with an AI professor you can chat, call, or meet "face-to-face". Engagement analytics for you.

persona approval engagement analytics live today at casealive.com
Try it here·casealive.com
casealive.com / makatemia / persona
A
Markus Tandefelt
CEO · Makatemia
"We made €588K last year. I have 25 people and one chance at the next product. Ask me anything about our business."
Chat
Call
Meet
§ 20Live moment · AI avatar Q&A
23 / 27
The tech you just saw, now pointed at me

Ask the
avatar.

Live questions, directed at my AI avatar. This is where office hours, tutoring, and student support are headed. The faculty who understand it early will shape how it's used.

Ask me anything
→ Ask anything about today's session
→ Try a topic from your own course
→ It also knows my research, teaching, and background
LIVE · Q & A
Mohammad Soltaniehha

Tap to start

§ 21Faculty exercise · Hands-on
24 / 27
§ 22Closer · Where AI is headed
25 / 27
Not just helping students · now securing the world's software

Project
Glasswing.

Anthropic-led coalition using an unreleased frontier model (Claude Mythos Preview) to find high-severity bugs in the open-source software the internet runs on.

1,000s of high-severity bugs, including zero-days in every major operating system and browser

One model found more critical vulnerabilities than most human security teams find in a year. Across every major OS and browser.

The coalition
Anthropic
AWS
Apple
Broadcom
Cisco
CrowdStrike
Google
JPMorganChase
Linux Foundation
Microsoft
NVIDIA
Palo Alto Networks
CLOSEThanks · Questions · Contact
27 / 27

Thank
you.

Mohammad Soltaniehha · Questrom · Boston University
casealive.com
Slides, tools, and case folder: email me for the link
QR code for slides
Scan for slides
soltaniehha.com/questrom