Working with AI · a field note for the team

You're already using AI. You're just standing on the bottom rung.

How one product manager went from chatting with an assistant to conducting a squad of agents — and how you can climb the same ladder.

Scroll to climb

Most of us use AI like a smarter search box. Ask, copy, paste. Useful — but you're still doing all the work. There's a ladder above that. Five rungs. Here's the climb.

Rung 01
Rung 01

One assistant, one task

You ask, it answers. We fed it months of community chatter and got the sentiment back in minutes — no dashboard, no infra, no analyst queue. Real value. But notice the bottleneck: every task still routes through you, one prompt at a time.

Mode: ask → answer
Rung 02

Give it hands

Connect the AI to your tools and it stops giving advice and starts doing the work. Wired into Jira, Metabase and Odoo, it queried over a million records, pulled live numbers across five databases, and wrote back — no copy-paste in sight. An advisor became an operator.

Mode: it reads & writes your systems
Rung 03

Give it memory and a workspace

A fresh chat forgets everything. So we gave it persistent memory and a shared board to work on. Now context survives between sessions — decisions, constraints, where we left off — and the work lives somewhere both of us can see. The AI stopped being a goldfish.

Mode: context that persists
Rung 04

From one agent to a squad

The leap. Instead of one generalist, a team of specialists: a lead that plans and reviews, a developer that writes code, a designer wired into Figma, a tester that breaks things. You hand out the cards; they execute in parallel; you review what comes back. You're not the doer anymore.

Mode: you assign, they deliver
★ The shift

From doing to conducting

Each rung trades a little of your time for a lot of leverage. By the top, the scarce skill isn't prompting — it's the conductor's craft: breaking work into pieces, giving sharp context, and judging the output. The AI got more capable at every step. The real upgrade was in how you work.

Mode: you orchestrate
The whole climb, in one line

Operator Orchestrator

You don't get a better AI at the top. You get a better way of working. The ladder isn't about the model — it's about how much of the doing you're willing to hand off.

Your first rung — this week

You don't jump to the top. You climb one rung. Adoption isn't a big-bang rollout — it's one task, one team, one rung at a time.

01

Pick a task

One repetitive thing you do every week. Boring is perfect. Hand it to the AI and judge the result.

02

Give it a tool

Point it at one system you live in — a sheet, a board, a database. Let it read before it writes.

03

Climb

Add memory. Add a second specialist. Each rung is small. The leverage compounds.