Over the past year, Lazer has embedded FDE teams inside some of the fastest growing AI companies in the world. Here’s 10 things we learned the hard way. Hopefully it saves you some time and $$. 1. Define KPIs before day one. No baseline = no ROI. Every engagement now starts with agreed metrics: tickets/week, time-to-first-PR, onboarding completion rate. 2. Agree on review SLAs. PR review delays were the #1# blocker. We now require a named reviewer, response SLA, and single access contact baked into the SOW. 3. First engineer builds the onboarding path. That documentation becomes the asset that makes every subsequent hire faster. Onboarding docs are part of the deliverable. 4. Schlep work is the moat. Access provisioning, environment setup, process gap filling, etc. This is exactly the work everyone else avoids. Once you’re the trusted pipe and intelligence layer, you’re very hard to rip out. 5. Calibrate talent one to two clicks above the end client’s team. Too far above and it creates friction. 6. Address the remote disadvantage proactively. In person onboarding week, inclusion in all standups, async response norms set upfront. 7. Track blockers quantitatively. Numbers surface what anecdotes hide. 8. Consequences ownership drives quality. Engineers who live with the decisions they make write better code. We push for post launch retainers on every engagement. 9. Map the real process, not the documented one. Run a workshop before scoping any agent workflow. The gap between those two is where the biggest value lives. 10. Build a reusable engagement template and let it compound. Each engagement funds the next. This is now standard practice across every FDE engagement Lazer runs. If your team is deploying agents and hitting friction at the enterprise layer, this is exactly what we’ve built for.