SXSW 2026 · Austin, TX
YOUR AI PROJECT
IS PROBABLY
FULL OF IT.
85% of AI projects fail* — not because of bad technology,
but because nobody can spot the bullshit before it's too late.
* Forbes / Gartner
Your Guide Through the AI BS
GREG NUDELMAN —
AI BULLSHIT DETECTIVE
Now Enrolling · Launching March 20
READY TO GET CERTIFIED?
UX for AI Professional Certification
A live, instructor-led 12-week program. One module per week.
Community support on Circle. By the
authors of UX for AI (Wiley, #1 Amazon New Release).
Full details at UXforAI.com.
12 live modules
Greg-led sessions
Circle community
ENROLL AT UXFORAI.COM →
Trusted By
Fortune 500 companies that hired Greg to find — and fix — the bullshit
CISCO
INTUIT
USAA
ROCHE
EBAY
SUMO LOGIC
The Full Case File
STOP SHIPPING BULLSHIT.
START SHIPPING PRODUCT.
Book, newsletter, certification program, and coaching — all at UXforAI.com.
GO TO UXFORAI.COM →
Case File · Exhibit A
6 SIGNS YOUR AI PROJECT
IS FULL OF IT
Every AI failure leaves evidence. These are the six signs that show up on every project that dies before it ships — in vendor pitches, internal roadmaps, and executive decks. Learn to read them before your budget becomes a crime scene.
🏎️
DEMO BULLSHIT
Looks incredible in a scripted demo. Falls apart the moment real users touch it with real data. Built to impress the board, not to ship.
📊
METRIC BULLSHIT
Accuracy: 94%! …on a cherry-picked benchmark. The model "works" — it just doesn't work on your data, your users, or your actual question.
🤖
CAPABILITY BULLSHIT
"AI-powered" plastered on a rules engine. Magic words — Agentic! Autonomous! — attached to a product that is none of those things.
🗓️
TIMELINE BULLSHIT
Production-ready in 2 weeks. Six months later you're still in "pilot." The estimate was made before anyone asked what data exists or what problem is being solved.
💸
ROI BULLSHIT
"10x productivity" with no baseline, no control group, no definition of what "productivity" means. A number invented to unlock budget.
🔮
ROADMAP BULLSHIT
A perfectly confident 18-month plan built on an AI system nobody has validated with real users. The further out the date, the deeper the BS.
Before & After
WHAT CHANGES WHEN YOU
CAN READ THE SIGNS
Spotting the signs is just the start. Getting certified means you know what to do next.
⚠ UNCERTIFIED — Can't Read the Signs
- ✗ Ships demos, not products
- ✗ Greenlit by insiders, never tested with real users
- ✗ ROI is a spreadsheet fiction
- ✗ "AI-powered" means an if/else statement
- ✗ Surprised by failure after 6 months and $2M wasted
- ✗ Lives in a terrified silo
✓ CERTIFIED UX for AI Professional
- ✓ Validates use case with real users before building
- ✓ Tests on thin slice: actual data, not sanitized samples
- ✓ Knows how to measure real ROI and risk from day one
- ✓ Builds agentic AI applications prototypes in hours
- ✓ Iterates fast: kills bad ideas in days, not months
- ✓ Trusted by leadership to lead AI projects
Case File · Probable Cause
4 SUSPECTS BEHIND
EVERY AI FAILURE
McKinsey, Gartner, BCG, and Forrester have all studied why AI projects fail.
Their findings converge on the same four suspects — systemic root causes
that appear in virtually every failed initiative, regardless of industry,
team size, or technology stack.
Our UX for AI Professional Certification was built to stop all four in thier tracks.
01
Suspect #1
WRONG USE CASE
Motive: Pressure to look innovative
The most common AI crime. Teams pick use cases because they're exciting, politically safe, or already on the vendor's demo reel — not because they solve a real customer problem with measurable value. The project passes every internal review and fails every user test.
McKinsey: 70% of AI projects fail to scale due to misaligned use case selection.
02
Suspect #2
WRONG DATA
Motive: Optimism about what exists
The model looks great on clean sample data. Then it hits production — incomplete records, inconsistent labels, missing history, and compliance walls nobody mentioned in the kickoff. The gap between "the data we have" and "the data we need" kills more AI projects than any algorithm failure ever will.
Gartner: Poor data quality costs organizations an average of $12.9M per year.
03
Suspect #3
WRONG ROI
Motive: Budget unlock at any cost
The business case was built to get approval, not to measure outcomes. Nobody defined what success looks like, what the baseline is, or who owns the measurement. By the time someone asks "did this work?" — the original sponsor has moved on and the numbers are impossible to reconstruct.
BCG: Only 10% of companies generate significant financial value from AI investments.
04
Suspect #4
WRONG PROCESS
Motive: Building before validating
Waterfall thinking applied to a probabilistic technology. Six months of engineering before the first real user sees it. No feedback loop, no thin-slice testing, no way to detect failure early. By the time the demo falls apart, the sunk cost is too large to admit the truth.
Forrester: Iterative, customer-validated AI development is 3× more likely to reach production.
The Verdict
THE CERTIFICATION
STOPS ALL FOUR
Suspect #1 Fix
The Storyboard Framing exposes the BS and finds the real use case worth solving — before a single line of code is written.
Suspect #2 Fix
The Digital Twin + RAG Registry maps exactly what data exists, what's missing, and how to close the gap.
Suspect #3 Fix
The Value Matrix builds a real ROI case from day one — defensible to the board, traceable to outcomes.
Suspect #4 Fix
The Snowball Sprint Framework is rapid customer-validated iterations using a lightweight AI POC you build in week one.
Case Closed
YOU KNOW THE SIGNS.
YOU KNOW THE SUSPECTS.
Now get certified to fix them. Full program, curriculum, and enrollment at UXforAI.com.
GET CERTIFIED AT UXFORAI.COM →
