The role of many faces

For a while in the startup world, I struggled to find a natural role for the varied set of skills and curiosity I had built up, with no clear functional expertise.

I never applied for a job posting, and I quickly learned that "strategy" roles rarely existed in early stage businesses. Pigeonholing myself into a corporate development or finance function at a later stage sounded like a path to boredom.

So I took on loose titles like Chief of Staff and VP of Operations & Strategy at Seed to Series A startups. Those titles gave me the flexibility to do cross-functional strategy and execution work, plus real ownership while most areas were going from 0 to 1.

The exposure in these roles is expansive. You figure out the unit economics of specific business lines, then learn those economics only hold up as well as the workflows, tools and market quirks underneath them.

The only way around those frictions was creative problem solving and constant iteration and testing, until there was enough structure to set up formal processes and move on to the next problem.

Enter Claude Code

These roles build a figure-it-out-at-all-costs attitude. Add the curiosity, risk-taking and light oversight that come with startups, and you can see why agentic tooling is the unlock for a cross-functional operator like me.

When I was introduced to Claude Code, I felt the last barrier had been lifted.

I started with the basic cross-functional workflows I never would have thought to tackle without engineers: consolidating invoicing and billing flows, and tying customer health signals to financial and growth health.

For one of my first uses, I turned a new deal process that once involved Slack, DocuSign, Google Suite, HubSpot and Stripe into one skill that ran the whole flow end to end in under 1 minute (down from 30+).

To get there, I had to map the workflow, add checkpoints for quality and accuracy, and monitor every run, since the process handled sensitive information for internal and external users.

Along the way, I learned how to access the endpoints of these applications, understand permissions and spot the exceptions, such as different contract structures that needed to be handled upstream. Catching those exceptions improved our future contracting process too.

And this was all before agents had their own virtual computers to run these same steps.

The only way to get good with these models is to learn the edges of the tools you're asking them to operate, and to have the access and curiosity to see how the process unfolds.

For those fortunate enough to be in similar positions, AI is both an executor and an educator. Learning the end-to-end build process that used to belong only to my technical counterparts let me stress test the assumptions of others, especially in engineering, product and core operations.

Every cross-functional operator knows one of the biggest barriers to alignment is not understanding how each function operates and where the real frictions are.

If an engineer tells me a task will take 8 weeks, and I can rebuild the building blocks in under 2 (not perfectly), I can at least get in the room to ask the right questions. More importantly, I can offer creative solutions, because the blockers usually come down to scope definition.

The Agentic Athlete

Before AI, cross-functional operators had visibility and access across the organization and could piece together why frictions existed. After AI, cross-functional operators can not only assess those frictions but act on them.

I've been calling this next generation of cross-functional operators the Agentic Athlete.

Agentic Athletes are part owners who tend to never own anything. They are technical, but not engineers. They are a mile wide, but deep where depth matters. Most importantly, they are curious, and they are becoming the AI-native culture carriers.

If this fits who you are or want to become, let's get in touch.

Originally published on X.