"Our industry isn't ready for AI." I've heard that one before.

I once ran a company that put autonomous shuttles on public streets in five cities. When we started, the technology wasn't fully baked. The legislation didn't exist. The market didn't exist. Riders were somewhere between skeptical and frightened. By any reasonable definition, nobody was ready, including us.

The company was Holo, and it grew into Europe's largest autonomous vehicle operator, seventy people across five countries. Not because readiness arrived. Because we stopped treating readiness as a precondition and started treating it as something you build in small, specific, permitted steps. We got politicians into dialogue instead of waiting for regulation to finish. We convinced cities to run bounded trials instead of pitching them a transformed transport network. We solved technical problems in the order the trials exposed them.

I think about that company every time a leader in pharma or biotech tells me their industry isn't ready for AI. Regulated, high-stakes, slow-moving by design: I understand the instinct. But "we're not ready" usually dissolves into three separate claims, and each one has a different answer.

"The rules aren't clear yet"

True, and they won't be clear for years, because the rules are being written in response to what companies do. Waiting for regulatory clarity before touching AI means letting other people's use cases define the rules you'll live under. The workable move is the bounded trial: one team, one workflow, data classes agreed in writing with your compliance people up front. At Holo we never asked a city to approve autonomous transport. We asked them to approve one route, with one shuttle, under named conditions. Small enough to say yes to, real enough to learn from.

"The technology makes mistakes"

So does every system in your building, which is why your industry already runs on review steps, four-eyes principles, and audit trails. AI slots into that discipline instead of replacing it: the tool drafts, a human who owns the output reviews. Choose workflows where a mistake caught in review costs minutes, not trust. Drafting, summarizing, first-pass analysis. Nobody's first AI workflow should touch a regulatory submission, the same way our first shuttle didn't drive a highway.

"Our people aren't ready"

This one is real, and it's also the most fixable, because readiness in people is mostly permission wearing a disguise. In the companies we train, capable professionals are already using AI quietly on personal accounts, waiting to hear it won't be held against them. The first thirty days of an open, sanctioned setup do more for organizational readiness than a year of committee meetings about it.

Readiness is downstream of starting

Here's what the shuttle years taught me. Every capability we became known for came out of a trial someone approved before we felt ready: the permits, the safety cases, the operating playbooks. The cities that said "come back when it's mature" got their shuttles years later, on terms shaped by the cities that started early.

Your industry will adopt AI. The only open question is whether your company's constraints and standards help shape how that looks, or whether you inherit the shape others settle on. One team, one workflow, thirty days. That's what starting looks like when nobody is ready.