- The Midas Report
- Posts
- 95% of AI pilots are failing and MIT just dropped the receipts
95% of AI pilots are failing and MIT just dropped the receipts
Here's the inside scoop on why it's happening
Midas AI
August 22, 2025
Cut through the noise, follow the real AI shifts.
Good morning from The Midas Report.
Todays newsletter is a BIG reality check, so if you’re ready, let’s get into it.
It turns out it’s not the models. It’s you.
That’s the uncomfortable truth buried in MIT Media Lab’s new “State of AI in Business 2025” report.
It found that 95%, yes, ninety five, of corporate AI pilots generate zero measurable return.
That means tens of billions in enterprise spend, much of it going to "innovative" use cases, ends up going absolutely nowhere.
The report looked at more than 300 AI initiatives and surveyed over 150 execs. While the headlines might tempt you to blame the usual suspects, ChatGPT, spotty data, stubborn regulators, the real culprit was more basic.
The culprit? Missing strategy.
Most companies are plugging in AI tools like they’re magic Legos, expecting transformation. Instead, they're getting PowerPoint decks and unanswered questions.
Most AI projects fail because they’re not linked to outcomes. No KPIs, no clear problem, no one with both the authority and the incentive to make it work.
And when tools are siloed in the IT department, tossed into sales workflows without integration, or hyped as “game changers” without a change in process, ROI isn’t just elusive.
It’s impossible.
The truth though? Success rate climbs massively when organizations bring in external partners and give implementation power to functional operators, not just tech teams.
Think “empowered accountable owner,” not “AI Center of Excellence.” When companies tie AI to real business systems, ERP, CRM, back office finance, impact gets real too.
Cost savings show up. Lead times drop. Risk goes down.
Sales and marketing might look shiny, but back office applications delivered better ROI in the study. And the winners weren’t always the biggest teams, they were the ones who mapped use cases first and built deliberately, not quickly.
So before tossing another AI SKU on the budget, ask a hard question, “is this solving a real business problem, or just checking a hype box?”.
🧠 The Download
AWS says replacing junior devs with AI is “the dumbest thing I’ve ever heard.”Amazon’s strategy is clear, don’t cut talent, compound it, positioning AI as leverage, not layoffs, to drive long term resilience in enterprise tech teams.
Vercel launches AI Gateway to stop your LLM stack from self destructing.
By solving cross model orchestration and failover at scale, Vercel just made running production AI apps way less fragile and protects teams from single provider choke points.
Google offers U.S agencies its full AI suite for just 47 cents a year.
This isn't charity, it's a land grab, Google is buying federal AI market share through cost leadership, tilting the B2G playing field before rivals can catch up.
Nvidia’s 3.2% stock drop ahead of earnings sent a chill through AI equities, punctuated by David Tepper’s selloff of key positions including Nvidia and Palantir, big name exits that read more like a warning shot than portfolio tweaking.
Autonomous delivery firm Nuro secured a commanding $203M Series E led by SoftBank, Baillie Gifford, and GV, proof that capital still flows toward AI startups solving hard, non glamorous problems with real world stakes.
Nexchain AI just pulled in $9.7M in presale funds, putting it on track as 2025’s top AI token breakout, promising decentralized, privacy first compute across blockchains.
That’s your daily AI edge for August 22nd 2025.
See you tomorrow!
Midas AI