The most underrated job in enterprise AI

What you should be doing over the next 18 months

At a founder workshop this weekend, someone asked what the most underrated role in an enterprise would be 18 months from now. I said it was the person who knows how to rebuild workflows for agents from scratch, and it sparked more conversation than I expected. Not the AI strategist or the prompt engineer, but the person who actually goes into an organization, understands how work really flows through it, and redesigns that work for a world where agents are doing meaningful parts of it.

I’ve been doing this hands-on for the past year, and what strikes me is how much genuine effort it requires and how rarely anyone talks about that honestly. Like a lot of things in life, the outcome gets all the attention while the setup gets none of it

Before an agent can do meaningful work inside any business function, someone has to make the unstructured data legible - years of documents, emails, notes, and tribal knowledge that lives nowhere an agent can read. Someone has to map the actual workflow, not the sanitized version on the org chart but the one that really runs, with all its workarounds and undocumented judgment calls baked in over years. Someone has to figure out which parts the agent handles well, where it breaks down, and where a human needs to stay in the loop. Someone has to connect systems that were never designed to talk to each other. And then someone has to rebuild the process itself, because the old one was designed around human constraints that no longer apply.

This is what the conversation in that room kept coming back to. Everyone is focused on what agents can do, and very few people are thinking seriously about the work required to make them actually useful inside a real organization. It doesn’t make for a good conference talk, and it won’t show up in anyone’s AI transformation case study. But it’s the work that determines whether any of the AI investment produces something real, and the people who can do it well are genuinely rare right now.

What I find most exciting is what this means for people early in their careers. This is one of those unusual moments where curiosity and willingness to get into the operational weeds matters more than seniority. The person who spends the next 18 months going deep on this - learning how to set up and redesign workflows for agents across any business function - is going to be valuable in a way that compounds pretty quickly.

Every organization is going to need people who can do for AI agents what good engineers did for software in the early 2000s: go in, understand the domain, and rebuild the infrastructure from the ground up. That wave created a generation of people who became indispensable fast, and this one will too.