Before you build, you see what a working team system looks like. We break down team AI operating systems CXL has found in the wild, what they have in common, and where the ones that fail went wrong.

• Breakdowns of team AI repos that work, and the structural patterns they share
• The common failure: a repo one person maintains and nobody else opens
• Outcome: A clear reference model for the repo you build in the next three parts

Your personal system assumes one user with full context in their head. A team repo cannot. This part covers exactly what to add: the folders, the skills, and the tools that make context shared rather than personal.

• Which folders, skills, and tools a personal setup is missing to work for a team
• How to move context out of your head and into files a teammate can read cold
• Outcome: Your personal library restructured as a team repo

A five-person content team and a 15-person growth org do not need the same structure. You build a folder structure sized to your team, your disciplines, and how your people work day to day.

• Structures for different team sizes, from a few specialists to a full growth org
• How to split folders by discipline, by workflow, or by both, and when each works
• Outcome: A folder structure built for your team rather than a generic template

The repo stays useful when changes get reviewed. You set up pull requests and branch rules so teammates can improve prompts and workflows without anyone pushing a breaking change to the team.

• Pull requests, branch rules, and reviewers set up for a marketing team, not engineers
• How to keep a QA layer that catches bad changes before the team runs them
• Outcome: A governed repo where anyone can contribute and nothing lands unreviewed

We open with the strongest AI systems we have seen running in B2B marketing, in-house and agency side, including the ones built in session one. You see what they share before anyone’s work goes under review.

• Breakdowns of the AI systems that hold up in production, and why they do
• What to reuse if you are starting from ours rather than your own
• Outcome: A reference model to judge your own structure against

Your structure goes in front of the room. Tycho Luijten reviews it as an outsider who would have to work in it, and the other attendees add where they would get stuck. Bring your own system or take one of ours.

• Expert review from Tycho Luijten, founder of the B2B agency Dapper
• Peer feedback from marketers and agency operators solving the same problem
• Outcome: A specific list of what breaks in your structure and why

The structure is half the problem. You map where AI adoption holds and where it stalls, across your own team or the clients you deliver for: which workflows run without you, who is comfortable, and which parts nobody touches.

• Where your team or your clients are comfortable with AI, and where they quietly avoid it
• How to tell an adoption problem from an execution problem
• Outcome: An honest adoption map, by workflow and by person, for your team or your accounts

You take the problems surfaced in parts six and seven and work through how to solve them, live. Approaches get proposed, pulled apart, and refined in the room until each gap has a fix you can run, whether the work lands with your team or a client.

• How to solve the gaps in your system, worked through live in the room
• What other teams and agencies tried against the same problem, and what came of it
• Outcome: An actionable plan to fix the adoption gaps the session surfaced