I have stopped calling myself the operator. I am the enabler now. An AI enabler, if I am being specific about it.
For years the chief of staff has been a force multiplier: someone who drives execution, gets leadership pointed in the same direction, and holds teams accountable to the objectives and key results they signed up for. All of that is human work, done by hand. I am enough of a believer in that cadence that I run my own life on quarterly outcomes too.
I think the seat is turning into a VP of AI. The execution and the coordination get played by automations running through agents, and the job of the person in the chair becomes bridging that transition.
The company brain
Most companies right now have people working with AI in silos, each one drawing only from their own documents. Where a shared system exists at all, it tends to be bureaucratic enough to slow everyone down, because it needs so much manual updating.
What I want instead is a source of truth that maintains itself. One on each customer, updating automatically every time there is a conversation with an agent listening in. One on how the company works, makes decisions, and gets things done, that can turn around and guide execution instead of only recording it. And one on the product that reflects how it actually works, so the enablement side of the business runs efficiently.
That is the vision, and I want to be straight about the distance between it and where we are. It is not happening yet. It is where I am leaning in over the next six to eight months, helping the company build a brain out of all our internal knowledge, something agents across non-technical roles will eventually be able to treat as a source of truth that is constantly self-healing and self-updating.
The first experiment
We started small at Amira. One or two people do all the enablement work here. They build the slide decks trainers use with teachers on how to use our product, and they hold the internal knowledge about things like how to request a custom report that shows student growth.
Step one was getting all of that into one place. It lived in five different locations, some of it authored by the CEO, some by the enablement people, some by others.
Then the actual work, with Claude Code and a set of agents. I am leaning on another employee, non-technical but fluent in Claude Code from using it for marketing, to help me run this first experiment. We are consolidating that enablement knowledge into markdown files and building an index that becomes the source of truth for how we enable our customers, at least on the customer success side. Sales is a separate problem.
I want to be honest about where this stands. The team is small, two or three people, and we have not yet taught the AI to update itself based on a conversation we have. That part does not work.
What does work: it can go and fetch anything we ask it to, and it can expose the gaps in what is out there for customer-facing people to use. It will say yes, I have that teaching guide, and the handout that goes with this slide deck, and the video of a child using that aspect of the product, here it is. And then: no, I do not have a video of a child using that other aspect. The gap is the useful part.
The hard part was never the tooling
A lot of that sounds technical and almost none of it is. It is people work, and team leadership work. The difficulty is getting the leaders in the room to make decisions about the heuristics and the rules the AI follows when it draws insights and updates the sources of truth.
That is where I tend to shine. Bringing people together to make hard decisions, and being the force that helps a team execute on whatever hard thing they are trying to execute on. It is the same work as building the customer success, sales operations, and talent functions from scratch, just pointed at agents instead of people.
The chief of staff was always the person doing that bridging by hand. As the agents take over the execution and the coordination, the bridging is the part that is left, and it happens to be the part I am best at.
