How I work with teams

I've been inside enough organizations to know this within a week of arriving: everyone can tell whether the leader actually uses AI. If you're not working with it and building with it yourself, no policy will convince your team to.

Leading this doesn't need to become your second job. I was CEO of a global health organization and grew the team from 20 to 70 people, so I know what it takes to lead change from the top. I also run my entire practice on an AI operating system I built myself, which means everything I show you is something I use every day.

In our first session I set you up with a lightweight AI system that knows you and your work, so it's useful from day one. After that, I use each session to work on your own fluency (lessons built around whatever is on your desk that week) and on your team: how they're using AI today, what worries them, and a plan for bringing them along, with the guardrails and privacy answers your people will rightly ask for.

Before I run anything, I talk with the leadership team and send everyone an anonymous survey, so I know exactly where people are: who's already using AI and how much, who's excited, who's worried, and what they're worried about. Most people turn out to be using it on their own already, without any shared agreement on what it's allowed to touch.

I tailor the two sessions to what comes back. We take the concerns people raised and work through them one by one, then define exactly what needs to be true for the team to feel comfortable using AI: what it's allowed to do on which work, at what level of review, what it never touches, and how your data stays protected. The security basics are part of that from the start, including safe settings on the tools, what Claude has to ask permission for before it acts, and which accounts and information stay off-limits. Your team writes all of it down as its own agreement, in its own words.

At the same time, I show the team what's possible with the work you already do, because the guardrails are there so people can be ambitious with AI. I want everyone to leave the second session excited about what they could build and clear on where the lines are.

In September I started Foundations with two teams. One of them, a global health organization, finished both sessions with a written agreement on how they'll use AI, drawn from the team's own words. For that team, Foundations was the first part of a longer program with me. Read how it went.

This draws on what I teach in TEMPO, my cohort for individuals, rebuilt for a team working in Claude together and shaped around how your organization works. Before we build anything, I spend time understanding your team's work: where the friction is, what people are trying to get done, and where AI would make the most difference.

In a public TEMPO cohort, everyone leaves with a system of their own. A private program for your team keeps that part (each person still builds on their own work, so what they build is in use the same week) and adds what the team shares: the instructions that let every person's Claude know your organization, shared skills for the work you repeat, and one standard for what it means to have used AI on something before it goes out.

By the end, your team has real skills and working systems they built themselves. They also have a shared way of using AI that no longer depends on the one or two people who were already ahead.

Every organization runs on a handful of processes it repeats constantly: hiring, procurement, forecasting, reporting, marketing. Most were never really designed. They're held together by manual work, a few key people, and knowledge that lives in someone's head, which is fragile and usually the reason things slow down as you grow.

I take one of those and rebuild it with AI in the loop, working with the person who owns it, so it runs reliably and stops depending on any one person. Your team ends up owning the process, the agents doing the heavy lifting, and the skills to keep improving it.

Some of what that's looked like: contractor management and procurement, financial forecasting, hiring, and marketing automation. See the case studies below.

You can also try one. I built a board view of a made-up nonprofit's finances that shows when salary money runs out, whether grants are being spent on time, and what happens if expected funding arrives late. It works the same way as the one I built for a nonprofit I work with, but every name and number in it is invented.

I join your team for six months, sitting in your team meetings and building with the people who'll use each system, so I know what everyone is working on. I work alongside the COO and the leadership team (someone else holds operations, policies, and compliance the whole time).

In the first month I mostly listen: I sit in the standing meetings, read what the team produces, and talk to each person about their work. From that I map where AI could take work off people's plates and where each person is with AI today, so we can decide together what to build first. I build that first system myself so something is running before the month is out.

From month two to month four we build, with each system made alongside the people who'll use it and a program for the whole team running at the same time. By month five your team is running everything with me still in the room, so problems come up while I can fix them. Month six is the handover (documentation, plus a named partner running the systems) before I step out.

I built the systems in the first two case studies below from inside those teams, as their fractional COO, which is the closest thing I've done to this engagement so far.

If six months is a lot to agree to before we've worked together, we can start smaller, with Foundations, executive partnership, or one discrete build. Then we'll both know whether the longer engagement is the right fit.