According to MIT’s latest report (via Fortune), a staggering 95% of generative AI pilots deliver zero measurable impact on the bottom line. That’s right—companies are sinking budget into “innovation theatre” without moving the P&L needle.

Now, I love AI. I use it in my work, my photography, even my wine content. But let’s be clear: AI isn’t a magic wand. It’s more like a Michelin-star kitchen knife—brilliant in the right hands, but you still need to know how to cook. Right now, most organisations are too busy admiring the knife to notice their soufflé has collapsed.

The core issue? Not the tech. It’s the learning gap. Businesses are shoving AI into workflows without rethinking culture, processes, or decision-making. It’s like giving your sommelier a chatbot and expecting it to pair wines better than they can. Spoiler: it doesn’t.

And yet, the 5% who get it right are usually scrappy startups. They pick one pain point, tackle it with laser focus, and scale with frightening speed. Zero to $20m in a year? That’s not just luck—it’s clarity, agility, and execution.

Meanwhile, corporates are busy pouring budget into sales and marketing pilots, when the real ROI is lurking in the back office: automating operations, trimming agency bloat, and cutting BPO dependency. Think less “shiny AI campaign”, more “goodbye, manual spreadsheet hell”.

The MIT report also makes another point I love: empower line managers, not just central “AI labs”. They’re closer to the pain, know what tools they’ll actually use, and won’t waste six months debating acronyms. Plus, vendor partnerships consistently outperform in-house tinkering—67% success rate vs a third as much when companies try to build their own. Translation: sometimes it’s smarter to hire the chef than to grow your own herbs.

Of course, all this sits under a cloud of exec worries—ESG, compliance, cyber risks, bubble fears. The usual suspects. But let’s not lose sight of the main course: if you’re not aligning AI pilots with clear business outcomes, you’re just feeding the hype machine.

For my world—wine, hospitality, premium brands—the lesson is obvious. Don’t ask “what can AI do for us?” Ask: “Which bottleneck could AI actually uncork?” Whether that’s personalised recommendations, smarter stock control, or cutting the hours wasted on reports, AI is only worth the pour if it leaves the business with a richer finish.

Otherwise, you’re left with a very expensive corkscrew, and nothing to drink.