Almost every leader I talk to right now is asking some version of the same question. Where do we put AI. Which function first. What tool. What does the rollout look like.Very few of them are asking the question underneath it, which is: what is actually slow around here, and why.I have spent thirty years running operations and owning P&Ls, and the last stretch of it building and operating several companies at once across recruiting, robotics service, home services, security, and investigations. Different industries, different economics, different customers. The pattern underneath is identical, and it has not changed because of AI. The constraint in most businesses is not capability. It is that nobody is sure who decides.
If your operating model is clean, AI makes you noticeably faster. If your operating model leaks, AI helps you leak at higher volume. It does not care which one you brought it.This is the part that gets skipped. Companies go looking for a use case when what they actually have is an ownership problem, and then they are surprised six months later that the pilot produced a lot of activity and no measurable change.PwC, Forbes, and most of the analyst coverage this year have landed in roughly the same place: a large share of enterprise AI programs have stalled, and the reasons are organizational rather than technical. That matches what I see. I have not once watched a serious AI effort fail because the model was not good enough.
Access gets mistaken for adoption. A company buys seats, sends an announcement, sees login numbers go up, and calls it done. Logins are not adoption. Adoption is when the output of a workflow is measurably different than it was before. If you cannot point at that, you bought software, not leverage.Nobody owns the workflow. This is the big one. The single highest-leverage thing I have done in every company where this went well was to name one person accountable for a specific workflow end to end, with the authority to change it. Analysts have started quantifying this and finding that organizations with a named owner get things into production at a dramatically higher rate. I believe it, because the alternative is a committee, and committees produce pilots.The process was never written down. If you cannot describe the steps on one page, including the exceptions that three people handle from memory, you cannot automate it. You will just encode the confusion and now it runs faster and you cannot see it happening.
Pick the workflow that costs the most and has the clearest edges. Not the most exciting one. The one where you can tell whether it worked.Write it down honestly. Including the ugly parts. The exceptions are where the value is hiding, because the exceptions are what the people are actually spending their time on.Name the owner before you name the tool. If nobody will take that job, that is important information about whether the workflow matters.Automate the narrow part first, measure it, then expand. Small and finished beats large and ongoing every time.None of that is a technology strategy. It is operations discipline, and it was the right answer before any of this existed. AI just raised the return on doing it.
The most common planning mistake I saw over the past two years was the assumption that you could keep the shape of the organization intact and retrain your way through. That is a comfortable thing to believe and it has mostly not held up. Roles change shape. Some of them change a lot.The leaders who lost trust were not the ones whose org charts changed. They were the ones who were vague about it for a year and then were not. People can handle a hard truth on a clear timeline. They cannot handle sensing something and being told nothing.If you are honest early, name what you actually know, and admit what you do not, you keep the people you need on the other side of it. That is worth more than any efficiency you gain in the transition.
This is the largest operating leverage I have seen in three decades of running businesses. It is also the fastest available method for industrializing a bad process. Both of those are true at the same time, and which one you get is not determined by the vendor you choose.The companies that come out of this well are going to be the ones that were already good at deciding things. If that is not your company yet, that is the project. The tools will still be there when you are ready for them, and they will be better.
Connect with Steve: linkedin.com/in/stevepurban