There's only one way to beat AI, and it's more AI. But the moment you adopt it, the models expose you.
Not your skills. Your process.
If AI doesn't produce speed on your team, typing was never the bottleneck. Decisions were. Every place approval lives in someone's head. Every spec that's a meeting nobody wrote down. If you can't write the skill, you never had a process. You had a person absorbing ambiguity.
At Ledger Collective, the person it exposed was me. I built the AI infrastructure — the data, the boundaries, the pipeline — and the work that was left changed shape. The scarce skill had moved to the accountant with the deep understanding of the work. And we're bootstrapped — every dollar of my salary was a dollar not going to the jobs we need right now. So I gave it back, and moved to the work that was left: sit next to the experts, teach, keep the data flowing.
The industry means the opposite when it says forward-deployed engineer: a vendor's engineer, rented, translating AI into your business. The translation layer, as a service. I mean deploy your engineers into your experts, until the experts are the builders and the engineers are the teachers.
The deal is simple. Two planes, two authors. Engineers own the data plane: governed, no side doors. Experts own the knowledge plane: skills, written in plain language. The AI does the labor in between, and a human stays on the record for everything that matters.
AI, meet me halfway. I'll build my half: the data. My experts will write theirs: the skills. You do the labor in between.
Build your half. Then get out of the way.
— Ashley
First shared on stage at the Capital One AI Speaker Series in Plano, Texas. Thank you, Howard Dierking, for hosting. One of four essays from that talk.
Keep reading Ship Smaller. With Proof. → We batched software into releases because shipping was expensive. It isn't anymore. Now batching is the risk.