“95% of AI pilots fail.”
“95% of diets fail.”
If I had a protein shake for every time I heard one of these claims, I’d bulk up in muscle mass the same way Claude Sonnet bulked up in tokens. Yet the first statement caused a panic, while the second barely earns a shrug. Both are built on the same overgeneralizations and contextual blind spots.
Recently, a study from MIT sparked panic in boardrooms and investor decks by declaring that a staggering 95% of AI projects don’t make it past the pilot phase. Cue the Fortune headlines, Wall Street indigestion, and shares of companies like Nvidia and Microsoft driven down. But here’s the twist: that number is as misleading—and misunderstood—as the infamous “95% of diets fail” stat that’s been passed around like gospel in the wellness industry for decades.
Let’s break this down, DAOFitLife-style.
Where the 95% AI Failure Myth Comes From
The MIT “State of AI in Business 2025” report made waves, but when you look closely, the methodology is much narrower than the headlines suggest:
- Period: January–June 2025
- Sample: Just over 300 “publicly disclosed” AI initiatives—meaning mostly press releases and SEC filings, not the thousands of internal pilots happening behind closed doors.
- Success definition: If a company didn’t publicly announce measurable productivity or P&L impact, it counted as failure. So if your AI project made employees’ lives easier but wasn’t in an earnings call? Failure.
- Sample size for interviews: Only 52 executive conversations and 150 survey responses—a thin base for such sweeping conclusions.
- Budget skew: Findings emphasized sales and marketing spend while overlooking engineering, software development, and agentic AI use cases actually driving adoption.
Commentators were quick to point out the clear fallacies of treating one study as a harbinger for the collapse of the AI bubble: see Anant US’s CEO and Intelcraft founder Rahul Singh’s post. It wasn’t about the tool — it was about the system. The MIT/NANDA report itself made clear the real problem was a learning gap: companies lacked the upskilling, workflows, and crack teams needed to translate AI into business results.
Similarly, the ‘95% of diets fail’ claim traces back to a tiny 1959 study of just 100 patients who were handed a diet and sent on their way—no coaching, no system, no accountability. That flimsy finding became ‘clinical lore’ for decades, even though modern evidence like the National Weight Control Registry shows thousands of people maintaining long-term success..
Buy, Build, or Partner: The Real Success Divide
And here’s another important detail: success rates differed dramatically depending on approach. Companies that bought AI tools from vendors had about a two-thirds success rate, while those that tried to build in-house succeeded only one-third of the time. That’s where strategic partnerships matter. The best implementations weren’t just a good tool—they combined vendor expertise with internal champions to create accountability, integration, and measurable outcomes.
Neither AI Nor Diets Fail—It’s the Way They’re Managed
The “95% of diets fail” myth comes from a 1959 study where 100 people were given a diet and zero guidance. No coaching. No system. Predictably, the majority failed—and that number stuck in headlines for decades. It discouraged people from even trying, and worse, it fueled an industry of shortcuts: crash diets, detox teas, and now the pharmaceutical quick fix—Ozempic.
But modern evidence tells a different story:
- In an 8-year study of 5,145 adults, over 50% lost at least 5% of their body weight.
- Of those who lost 10% or more, 65% kept it off long-term—with the right coaching, systems, and accountability in place.
So it’s not that diets inherently fail. It’s that the management of the diet process fails when people go it alone, without structure, support, or expertise. The ones who succeed rarely do it solo—they build partnerships: coaches, trainers, nutritionists, communities. In other words, they assemble a crack team to keep them informed, accountable, and on track.
And this is where the parallel with AI becomes sharp. As Rahul Singh put it: the solution isn’t in AI alone, or in people alone. It’s in the management of both. AI projects aren’t IT projects. They’re reconstructions of how a company works—requiring leadership that can orchestrate people, processes, and technology into a system that delivers more value than the sum of its parts.
The failure isn’t the model. The failure is in skipping the hard work: building literacy, designing workflows, setting up incremental wins, and assembling the crack teams and strategic partnerships that make it stick.
In both fitness and AI, you don’t blame the tool. The tool is good. But without management, education, and the right system of accountability, it will never work.
The Parallel: AI Pilots = Crash Diets
Both AI deployments and diets fall into the same traps that can cause them to go off track:
| Trap | AI Pilots | Diets |
|---|---|---|
| No guidance | Drop a model in Slack and pray | Follow an app with no coach |
| Unrealistic expectations | 10x productivity in 2 weeks | Lose 20 pounds in 20 days |
| Misaligned incentives | Measured by press wins, not adoption | Measured only by pounds lost |
| Lack of integration | Not embedded in workflows | Not embedded in lifestyle |
| Quick fixes | Vendor hype cycle | Ozempic, crash diets |
| Lack of partnerships | Ignoring vendors/experts | Refusing coaching or community support |
AI or diets, the trap is the same: no reliable system and a user experience so clunky that no one sticks with it.
How to Make It Work (For AI and Fat Loss)
Whether you’re rolling out an LLM or trying to lose body fat, here’s the shared secret:
Don’t do dumb stuff. Build the system, not just the tool. And get a crack team—and the right partnerships—to keep you accountable.
For diets (or better yet, diet habits), that means:
- Stop relying on “cheat meals” (aka scheduled binges). If it feels so restrictive you need a break, it won’t last.
- Create a realistic calorie deficit: more protein, fruits, and vegetables; swap soda for diet soda; limit eating out; choose leaner proteins and lower-fat dairy.
- Don’t cut out foods you love. Use an inclusive mindset instead of an exclusive one.
- Follow the 80/20 rule: eat nutritious, minimally processed foods most of the time, but leave space for enjoyment.
- Strength train 2–4 times a week to protect muscle while losing fat.
- Walk daily—8,000+ steps beats sitting all week after a couple of gym sessions.
- Prioritize sleep (7–9 hours). Sleep-deprived people eat more, period.
- Monitor progress with multiple metrics: weight, measurements, photos, and workout logs.
- Build accountability and surround yourself with a crack team: a trainer, nutritionist, supportive friends, or a positive gym environment. These strategic partnerships make the difference between short-term effort and long-term results.
For AI pilots, the same logic applies:
- Don’t rely on gimmicks.
- Design projects with sustainable workflows.
- Integrate tools into daily habits, not just one-off demos.
- Measure success with multiple indicators, not just P&L headlines.
- Build accountability with internal champions, and lean on strategic partnerships with vendors and experts. The best results happen when you combine external expertise with internal ownership.
Final Rep
The “95% fail” myth sounds scary—until you realize it’s mostly noise. AI isn’t failing. Diets aren’t failing. What’s failing is our habit of throwing tools at complex problems without building systems—and teams—around them.
Instead of asking, “Will it work – or will it fail?” ask a better question:
What support, structure, and system are you putting in place to make it stick? And who’s on your crack team or strategic partnership bench to keep you accountable?
Because hype doesn’t scale. But habits—and partnerships—do.
