How AI pays off

The now-famous MIT study showed 95% of generative AI pilot programs produce zero measurable impact on the bottom line. 95%. That’s MIT, not a vendor with an agenda. In the majority of cases, the ROI isn’t absent because the technology failed. Root cause analysis shows nobody did the correct diagnosis. The business pain was vague. The scope was a wish list. Nobody defined clearly what success would look like. And nobody planned for what happens after the demo stops being exciting and the system has to work in real-life conditions, every day, for the duration.

This is where companies get hurt. The vendor shows a beautiful demo. The system goes live. And then… nothing. No way to tell if it’s working as advertised. No humans reviewing the output. No benchmarks to evaluate the performance against. No one accountable. Lights are on but no one’s home. AI does not work like ordinary software. Fundamentally, the brains behind your AI system aren’t built like normal software. They are grown like a plant. You can send the same input 1,000 times and get 1,000 slightly different outputs. It can start giving worse or wrong answers over time without throwing an error. 3 months later, you get a harmful answer or a deal falls through, and everyone’s surprised. They shouldn’t be. Nobody was watching. And that is all assuming you selected one of the better vendors in the first place.

What we believe

Companies that are already getting substantial quantifiable business value from AI today have one thing in common. They did not start with some grand transformation narrative to reach operational maturity levels that would make Accenture proud. This is not something you’re likely to hear from anyone else, but unless you’re a Fortune 500 company, you’re very probably not ready for wholesale “digital transformation” through AI.

The companies we help succeed are finding one specific problem that is isolated, repetitive and clear — follow-up is inconsistent, intake takes too long, knowledge is trapped in two people’s heads — and they’re picking one associated workflow or system to fix and one point solution with which to fix it. Low-hanging fruit with what we like to call micro-ROI. Easy consensus on what success looks like. Prove it works there. Measure it. Enjoy the ROI. Then expand. The goal is not to bet the company on AI. The goal is to find the first few easy wins.

People with superior tools

Here’s a fact that surprises people: 82% of small businesses using AI actually grew their headcount. They hired more people. They did not start a wave of layoffs. Because when AI handles the low-value repetitive work, when AI breaks apart a job description into component tasks and assembles them in a different order, the operating model changes for the company, and the team has capacity to take on more clients, more revenue.

The job destruction narrative is out there, and it is not unfounded. Public transit terminals in the Bay Area have full-size ads that say: “Stop hiring humans”. But as of today, we believe that using AI to cut staff as many large corporates have already done is the wrong end of the binoculars.

What’s the right end? Why do the same with fewer people if you can 5–10x your output or value to clients with the same staff? Only 8% of workers say they want AI running the show. A future with mass job destruction may or may not be inevitable. For the duration, we can make this about people with superior tools massively outperforming people without them.