Team EnablementLadder.
AI changed what individual engineers ship in a week. Most teams have not changed what they expect of each other, how they pair, or what “senior” means. The capability shift moved the bottleneck. The ladder moves the team.
Why Now
The bottleneck moved. The team didn't.
The capability shift
Individual engineering output is unrecognizable from two years ago. A junior with a working agent setup ships features that would have been senior work in 2023. The gap between top and bottom of the team widens fast, in both directions: the top gets further ahead, and the bottom is no longer where it was.
Most teams notice this as a vibe. Some notice it in the calibration meeting. Almost no team has a deliberate ladder for it.
The team-shaped bottleneck
Hiring bar from 2022
The interview loop still tests live coding without tools. The team works with tools.
Review rituals from 2022
PR reviews scan diffs. The diff is no longer the unit. The spec and the eval are.
Senior definition from 2022
Promo committees grade against the old bar. Quietly, the team is capped at last year's ceiling.
Principles
Six ideas the framework rests on.
Capability is a team property, not a person property.
Teams ship. Individuals contribute. A team can be operating two ladder levels above its most senior individual if the rituals are right. The reverse is also true.
Ladders are exits, not gates.
Each level is a step out of the previous one. You graduate by leaving the rituals of the level below behind. Treating levels as permission slips creates queue and resentment.
Tools without rituals do not stick.
Copilot adoption is a leadership artifact, not a procurement decision. Buying seats and waiting for adoption is the most expensive way to learn that you needed rituals.
Apprenticeship beats workshops.
The fastest way up the ladder is shadowing someone one level above you for a week. Six-hour workshops produce slides, not behavior change.
Measure adoption, not enthusiasm.
Slack threads about AI are not AI in production. Track the artifacts: specs written, evals shipped, agents merged. Enthusiasm is a vanity metric.
Senior gets redefined every 18 months.
The capability bar moves. So does the title definition. Promo committees calibrating against the 2023 senior bar are quietly capping the team.
The Ladder
Five levels. One direction.
Each level names a different relationship to AI in the work. Read the level. Place yourself. Place the team. The next-level indicator is the cue, not the credential.
Level 1
AI-Curious
10%
Has tried Copilot or ChatGPT. Does not yet use it for daily work. Skeptical or unsure where it fits.
Level 2
AI-Comfortable
30%
Daily user. Single-step prompts. Treats AI as a smarter autocomplete. Rarely structures work for AI consumption.
Level 3
AI-Confident
55%
Multi-step prompts. Knows what AI is bad at and routes around it. Begins to write specs and tests with the agent in mind.
Level 4
AI-Fluent
80%
Treats AI as a teammate. Builds and uses agents, evals, and custom tools. Designs work so the agent does the parts agents are best at.
Level 5
AI-Native
100%
Builds the rails the team rides. Designs evals, owns model strategy, sets adoption metrics. Operates one layer below the team.
Newsletter
Want the next chapter when it drops?
Short dispatches between chapters. Field notes, not summaries.
The Framework
How the ladder runs in practice.
Six pieces. Assessment, the pilot program, the rituals, tooling, manager moves, and anti-patterns. Each one stands alone. Together they make the ladder operate.
Placement is two passes: self-assessment first, manager calibration second. Both pass through the same rubric: what they ship, what they teach, and the next-level indicator. Self-assessment alone overweights enthusiasm. Manager assessment alone underweights work that happens between standups.
Signal
- · Specs written in the last 30 days
- · Evals shipped in the last 90 days
- · Spec-to-merge time on the engineer's work
- · Pairing sessions led, not attended
Noise
- · Copilot acceptance rate in isolation
- · Slack messages about AI
- · Workshop attendance
- · Self-reported confidence
Recalibrate every quarter. The bar moves faster than promo cycles, so the assessment cadence has to lead the calibration cycle, not lag it.
The first run of the ladder is a cohort, not a memo. Pick 8 to 12 engineers from one team or two adjacent teams. Mixed levels by design: at least two engineers at each level the cohort spans.
Days 1 to 30
Pairing-heavy
Daily pairing arcs across adjacent levels. Pattern: 30 minutes, one ticket, one observation in the team channel.
Days 31 to 60
Ritual install
Friday demos start. Eval clinic in week six. First adoption-not-enthusiasm review at day 45.
Days 61 to 90
Sponsorship arcs
Each Fluent or Native picks one Comfortable or Confident to sponsor. Day 90 closes with a public placement review.
The pilot succeeds when the rituals continue after day 90 without leadership reminders. If they stop, the pilot installed enthusiasm, not capability.
Four rituals carry the ladder. None of them are workshops. All of them are time-boxed and recur on the team calendar.
Daily AI pairing window
A standing 30-minute slot on the team calendar. Anyone can drop in with an in-progress ticket and pair with whoever is in the room. Default partner is the level above the driver.
Weekly Friday demo
Five 8-minute demos, each one showing a single move up the ladder this week. Not features. Moves. The audience is the team, not the org.
Monthly eval review
The team reviews every eval shipped in the last month. What broke. What it caught. Which one to retire. Owner of each eval rotates.
Quarterly anti-pattern callout
The team names the anti-pattern they see in their own work. Then they name the one they see in adjacent teams. The exercise is short, public, and required.
Tools without rituals do not stick. So tools follow rituals, not the other way around. Each level gets the toolset that the level's rituals require, and nothing more. Overprovisioning a Comfortable with a Fluent toolkit slows them down. They spend the week on the tool instead of on the work.
| Level | Table stakes | Wait on |
|---|---|---|
| Curious | Chat IDE plugin. One model. | Agents, custom tools. |
| Comfortable | Chat IDE plugin. Inline diff suggestions. | Agent orchestrators. |
| Confident | Two models for switching. Spec templates. | Multi-agent setups. |
| Fluent | Agent runtime. Eval harness. MCP-ready tooling. | Custom platform work. |
| Native | Everything. Owns the team's tool selection. | Nothing. |
Managers operate the ladder by changing what they do at each level transition. The work changes more than the title does. The list below names the manager move that unlocks the next level for the engineer.
L1 → L2 Sit in on three pairing sessions. Sponsor the engineer into the daily pairing window. Do not assign training.
L2 → L3 Ask the engineer to author one spec a week. Review it like a senior peer would, not like a manager. Stop sending tickets without context.
L3 → L4 Co-design one agent workflow with them. Hand over the eval that catches its failures. Make them present it.
L4 → L5 Hand them a platform problem. Step out of the room. Show up at the readout.
The Sheep Dip Workshop
A full-day session that produces slides and no behavior change. Six hours of attention for two weeks of nothing.
The Lonely Champion
One engineer carries the entire AI adoption. When they leave or burn out, the program goes with them.
The Surveillance Metric
Tracking Copilot acceptance rate as a performance signal. The team learns to accept noise to look busy.
The Adoption Theater
Loud Slack channels about AI usage, no change in spec-to-merge time or eval coverage. Performance, not adoption.
The Tool Spree
Procurement-led rollout. Five new licenses, no ritual. The seats get assigned. The tools sit idle.
The Forever Pilot
A pilot that never ends. No day 90 review. The cohort fragments back into the team without a placement update.
The Senior Skip
Treating seniors as exempt from the ladder. They were calibrated on the old bar. They need the ladder more, not less.
The Junior Trap
Loading juniors with Fluent-level tooling to fake productivity. They learn the tools and not the craft. Capability looks like it grew. It didn't.
Sample Artifacts
Three artifacts the ladder actually produces.
The ladder is not a slide. It is the artifacts that come out of running it. Three examples below.
Onboarding plan for an AI-Confident engineer joining the team
# 90-Day Plan: Senior Engineer, AI-Confident at Hire ## Context New hire enters at the AI-Confident level. Strong on prompt design, weaker on agent design and team rituals. Goal at day 90: operating fluently inside the team's pairing and eval rituals, sponsoring one Confident-to-Fluent move on the team. ## Days 1 to 14: Land **Pairing** - Pair daily with the team's most senior AI-Fluent engineer for the first ten working days. - One ticket per day, picked from the bug rotation. Pairing time-boxed to 90 minutes. **Reading** - The team's spec library, two specs per day. Skim the maturity-model placement notes. - Last quarter's eval review notes. Identify the three evals most relevant to the new hire's surface area. **Output** - Ship two small fixes paired. Write one spec for a change they would have shipped without one. Pair-review the spec. ## Days 15 to 45: Calibrate **Pairing** - Reverse the direction. The new hire pairs as the driver. The Fluent engineer reviews and points out the rituals being skipped or over-applied. **Specs and evals** - Author one full feature spec end-to-end. Submit to the team's spec review channel. - Co-author one eval with the engineer who owns the surface area. Ship it. **Manager moves** - 1:1 every Tuesday and Friday for the first month. Drop to weekly after. - One direct calibration session with the team lead at day 30. Confirm placement, surface any signals of next-level readiness. ## Days 46 to 90: Sponsor **Cohort role** - Pick one Confident-level peer. Run a four-week pairing rotation aimed at the Confident-to-Fluent transition. - Co-design one new agent workflow the team will adopt or reject. **Output** - Ship two medium features as the spec author. The agent does the implementation pass; the new hire owns the eval gates. - Lead one Friday demo on a pattern they have brought from prior experience. **Day 90 checkpoint** - Adoption review with the team lead. Confirm fluency, name the next 90-day arc. - Update the team's hiring rubric with one observation from the new hire's first 90 days.
Keep reading
Next chapter in the newsletter.
Short dispatches between chapters. Field notes on what teams actually run, what gets dropped, and the rituals that stick.