5 Levels of AI Adoption.
Which one are you on? Prove it.

Where your engineering org actually is. What is required to be there. What to measure to prove it. What to stop doing this quarter.

The Rule

5 levels. You cannot skip.

95% of AI pilots never reach production. The reason is almost never the model. It is the skip. Companies jump from L1 to L4 without doing L2 talent or L3 measurement. The dashboard looks good for two quarters, then someone asks what the AI line on the P&L bought us, and no one has an answer.

Below: each level, what is required to be operating at it, the 2 or 3 KPIs that prove you are there, the red flag failure mode, and the next move to climb.

The Trajectory

Three KPIs grow with you. Tap a level to see where you should be.

Code touched by AI. Engineers fluent enough to ship with it. AI on the P&L. All three climb. The P&L line lags. That is the L3 trap.

100% 50% 0% L1 L2 L3 L4 L5
AI in the code
70%
of PRs touched by AI. Measured monthly, tracked against delivery stability.
AI-fluent engineers
80%
can ship a feature with AI assistance. Audited quarterly with a real demo.
AI on the P&L
4%
attributable margin from AI. The number the CFO signs.
1
Level 1

Nothing in Production

Hackathons, pilots, one engineer using Copilot on her own card. No paying user has triggered an inference. No AI line on the P&L.

KPIs to track

KPIRedAmberGreen
AI-assisted PR rate <10% 10 to 30% >30%
Days to first production AI feature Never set Slipping <90 days
Paid AI seats deployed <25% 25 to 60% >60%

What is required to leave L1

  • One workflow chosen. Internal, bounded blast radius. Not customer-facing on day one.
  • Baseline measured before any AI lands. Lead time, tickets, cost per unit. Numbers a CFO can read.
  • Six-week timebox. If it does not move the baseline number, kill it visibly.
  • One team owns pilot and production. No throwing over the wall.
Failure mode

"We have 14 pilots running." No, you have 14 ways to stay at L1.

Compliance and security used as the reason for not shipping. Sometimes legitimate. More often a polite stall.

2
Level 2

Talent

Your org chart changed because of AI. Not your tooling. Your org chart. Headcount reshaped. A named owner runs AI enablement with budget.

KPIs to track

KPIRedAmberGreen
% engineers AI-fluent, audited quarterly <30% 30 to 70% >70%
Net AI-driven org change, last 12 months 0% 5 to 15% >15%
Single named AI-enablement owner No In name only Yes, with QBR

What is required to be at L2

  • Job descriptions list AI fluency as baseline. Not a bonus. A requirement.
  • Headcount allocation has visibly shifted. Fewer junior backfills, more AI-fluent senior hires. Or vice versa.
  • At least one role was eliminated or repurposed. Because AI now does that work.
  • One named AI-enablement owner. With a quarterly business review, not a Slack channel.
Failure mode

Layoffs without reskilling. Brand damage now, hiring problem in 18 months. Companies that layoff then rehire spend 1.4 to 2x more over 24 months than companies that reskilled.

Treating AI fluency as a single training course. Real fluency is daily use, not a half-day Zoom.

3
Level 3

Proof

You can answer one question with one number. What did AI cost us last quarter, and what did we get. If CFO and CTO disagree, you are not at L3.

KPIs to track

KPIRedAmberGreen
AI cost per shipped unit, tracked monthly Untracked Tracked, no threshold Tracked, threshold at QBR
AI-attributable margin Untracked <2% >2%
AI delivery instability ratio vs non-AI 2x+ worse 1.2 to 2x Within 20%

What is required to be at L3

  • AI is a named P&L line. Owned by Engineering or Product. Not buried in a cloud bill.
  • Unit economic exists. Cost per AI-assisted PR. Cost per AI-resolved ticket. Per unit of value shipped.
  • Counterfactual baseline. What it would have cost without AI. No baseline, no proof.
  • Delivery stability tracked alongside throughput. DORA 2024 and 2025 are clear: AI lifts throughput, often drops stability. If you only measure one, you miss the trap.
Failure mode

Counting Copilot seats as ROI. Seats are an input. Output is what shipped.

50% of leaders want business outcomes. 3% measure them. That gap is where most of the audience sits.

Token costs dropped 98% in 24 months. If your monthly AI bill stayed flat, you are not capturing the deflation. Someone else is.

4
Level 4

Going AI-Native

AI is not a tool your engineers use. It is a teammate in the build pipeline. Task complexity rising. Autonomy rising. Human turns per session falling. All three at once.

KPIs to track

KPIRedAmberGreen
AI-merged PR rate, fully autonomous <5% 5 to 30% >30%
Production AI features with CI eval coverage <40% 40 to 80% >80%
AI-attributable EBIT, signed by the CFO 0% <5% >5%

What is required to be at L4

  • Customer-facing AI feature with an SLA and on-call rotation. Not a demo. A product.
  • Autonomous agents handle a measurable share of internal work. Merges, ticket resolution, security patching.
  • Eval harness in CI. Regressions block deploys. Hallucination and safety monitors page the on-call.
  • CFO signs her name to an AI-attributable EBIT number. Not directional. Specific.
Failure mode

No eval harness. Klarna shipped customer-service AI in Feb 2024 and partially reversed in May 2025. Volume metrics looked great. CSAT and NPS on edge cases collapsed. The team was not measuring quality, only throughput.

Declaring "we are AI-native" in marketing before saying it in the engineering all-hands. If your team does not say it on a Monday, you are not.

5
Level 5

You've Won the Race

Your competitive advantage is no longer "we have AI". It is "we ship a product that could not exist without our specific AI maturity". Rare. Probably temporary.

KPIs to track

KPIRedAmberGreen
% revenue from AI-dependent product capability <5% 5 to 20% >20%
AI-native hiring premium vs market 0% <10% >10%
Net-new AI-native bets in roadmap, with owners 0 1 to 2 3+

What is required to be at L5

  • Customer-facing brand moat that competitors cannot buy. Not the model. The capability around the model.
  • Net-new product lines exist because of internal AI maturity. Shopify Sidekick. Klarna shopper assistants. Cursor IDE itself.
  • You hire talent others cannot. Because AI work here is more interesting than AI work elsewhere.
  • Velocity has changed category. Not 20% faster. A different kind of work shipped.
Failure mode

Calling yourself AI-native because the website says so. Marketing the level you wish you were at is the most common L4 self-deception.

Treating L5 as an end state. Anthropic's internal data shows the complexity and autonomy curves still rising. There is no plateau in 2026.

The Journey

0 to 5. Where you are. Where you go next.

  1. 0
    Live in sprints
    0 paid AI seats
  2. 1
    120 Pilots
    0 production AI features
  3. 2
    Few advocates
    >70% engineers AI-fluent
  4. 3
    Prove it!Measure impact on product
    >2% AI-attributable margin
  5. 4
    AI-Native
    >5% AI-attributable EBIT
  6. 5
    Shine
    >20% revenue from AI product
Engineering Leaders Community