The Bottom Line Up Front
MD Anderson spent $62 million on IBM Watson. Zero patients treated.
Zillow lost $500 million on its AI algorithm. 2,000 jobs cut.
DOJ is now suing RealPage and six landlords for AI-enabled price-fixing affecting millions of renters.
Your stalled AI project? It’s heading in the same direction unless you fix three things this week.
Why Your AI Project Is Actually Stuck
Every week, someone tells me: “We’re six months in, spent $200K, and nobody’s sure if it’s working.”
Here’s what they discover: The technology isn’t the problem. The alignment is.
Over 80% of AI projects fail, which is twice the rate of regular tech projects. But they don’t fail because AI doesn’t work. They fail because:
Three Disasters To Ponder Before Your Next Build
IBM Watson at MD Anderson: $62M, Zero Patients
The plan: AI-powered cancer treatment recommendations that would revolutionize oncology.
The reality: Five years later, $62 million spent (original estimate: $2.4 million), and the project was shut down before treating a single patient.
What killed it? Shifting project focus mid-stream, clinicians not involved in development, and complete misalignment on what success meant. Full story here
Zillow Offers: $500M Loss, 2,000 Jobs Gone
The plan: AI algorithm buys and flips houses at scale.
The reality: The algorithm couldn’t keep up with market changes. Result: $500+ million in losses and 25% of staff laid off.
What killed it? No proven manual process to automate, ignored frontline signals about market conditions, and an algorithm solving yesterday’s problem today. Full story here
DOJ vs. RealPage: Algorithmic Collusion Lawsuit
The plan: an AI algorithm helps landlords optimize rental pricing across millions of units.
The reality: DOJ sued RealPage in August 2024, alleging their algorithm enabled price-fixing that harmed millions of renters. The software collected nonpublic pricing data from competing landlords, fed it through a common algorithm, then recommended rents, essentially enabling collusion disguised as “optimization.”
What’s going wrong? The algorithm replaced competitive market dynamics with coordinated pricing. Six major landlords are now defendants, and Greystar (managing 950,000 apartments) has already settled to stop using the software.
Your AI Project Unstuck Checklist
If your project is stuck right now, answer these three questions:
✅ Can you define the manual process being replaced?
- Ask: What specific task are we automating? Who does it today? How long does it take?
- Red flag: If you can’t describe the current process in 2-3 sentences, you’re not ready.
✅ Did you pick a fast, painful use case?
- Ask: Does this cost money every day? Can we measure results in 60-90 days? Will operators actually want this solved?
- Red flag: If results take 6+ months, it’s too big for a first project.
✅ Can you get an early win to build trust?
- Ask: What’s the smallest useful version? Can we pilot with 5-10 users first? How will we celebrate wins?
- Red flag: If your plan doesn’t show value until month 12, nobody will stick around.
What Winning Actually Looks Like
Not all projects start with “digital transformation.” We recently worked with one of the largest construction rental companies in the U.S.
They started with: “Customers keep calling with basic troubleshooting questions. Let’s fix that.”
- Month 1-2: Built Gen AI troubleshooting assistant
- Month 3-4: Rolled out to customers
- Month 5: Call volume started dropping
- Month 6: 2M+ call minutes saved annually
Time to ROI: 6 months.
Then they expanded. But they started with concrete, painful, and measurable.
The Getting-Unstuck Action Plan
This week:
- Schedule a 2-hour “reset” meeting with leadership, operators, and the implementation team
- Answer three questions together:
This month:
- Cut scope: Pick one painful problem, ignore everything else
- Shadow operators: Watch how work actually gets done
- Build 90-day milestones: Define “working,” not just “launched”
This quarter:
- Get one measurable win
- Document what worked
- Use that win to fund the next problem
The Real Truth
Your AI project isn’t stuck because AI doesn’t work.
It’s stuck because nobody agreed on what “working” means, operators weren’t involved, and leadership is solving a different problem than the frontline.
Fix these three things, and your project starts moving again.
The companies winning with AI aren’t the ones with the biggest budgets. They’re the ones who picked one specific problem, got everyone aligned, and actually solved it.
Then moved to the next one.
