Self-Promotion
Show off what you're working on
Start new thread
Lina Huchok

MIT Report: 95% of GenAI pilots in business flop. What do you think about this?

by•

Hi, PH community!
MIT’s Project NANDA just released The GenAI Divide: State of AI in Business 2025.

And the numbers are… brutal:

  • 95% of GenAI pilots fail to deliver real results.

  • The money and hype are there, but ROI? Zero.

  • Only 5% succeed, and those are the ones laser-focused on solving a clear business problem.

  • Biggest ROI right now? Back-office automation (HR, docs, workflows), not flashy sales or marketing tools.

  • External providers win, partner-led projects are 2x more successful than internal experiments.

  • Shadow AI is everywhere, employees use ChatGPT & co outside official policies. Risky? Yes. Opportunity? Also yes.

Bottom line for founders & marketers:

AI is not a magic button. It’s process + culture. Those who integrate quickly, in small profitable steps, are the ones who win. What do you think?

62 views

Add a comment

Replies

Best
Valeriy Yasakov
Really interesting insights! I wonder: 1. When they say 95% of pilots “fail”, is that mostly about not scaling or truly negative ROI? 2. And for the 5% that succeed - what do they usually do differently at the very start? Thanks!
Oleksandr Samoilenko

@lina_huchok Totally agree with these insights. Most companies still launch GenAI pilots for the hype, not to solve a measurable business problem. That’s why ROI shows up mainly in structured processes like docs, HR, or support.

Also spot on about partnerships: external expertise helps avoid beginner mistakes and speeds up results.

Shadow AI is a double-edged sword—yes, it’s risky, but it’s also proof that employees already see value. Better to channel that into controlled adoption than to block it.

In my view, the next real wave isn’t generic automation, but narrow, function-specific GenAI tools that go deep into core workflows.

Lina Huchok

@oleksandr_samoilenko Thanks, Oleksandr, totally agree with you, especially on shadow AI. It’s often treated as a threat, but in reality it shows employees are already hungry for better tools. The challenge is to transform that “underground adoption” into structured, value-driven use cases.

I like your point about the next wave being narrow, function-specific GenAI. Curious, where do you think this shift will start showing the strongest impact first: in knowledge-heavy industries (like law/finance) or more operational areas (like logistics/supply chain)?