Find where the work piles up before you pick an AI playbook
Four engineering organizations put AI on four different problems. What transfers from them is the diagnostic they all ran first: where does the work actually pile up?
Written by
-
Benedikt Stemmildt Founder & Co-CEO
Published: 20 August 2026 Updated: 2 min read
I spent the morning re-reading the AI case studies from bigger engineering orgs. Four of them line up in a way I didn’t expect.
Four companies, four problems
A wholesale marketplace company put AI on the code review path: an internal reviewer that handles about 3,000 code reviews a week across its entire codebase (opens in a new tab).
A social platform company put AI on its existing test suite. The agent proposes improvements to tests humans already wrote, and engineers accepted around 73% of what it proposed. The slow work of hardening a test suite stopped piling up.
An investment bank put AI on legacy-code migration. The agent reads decades-old code and writes a plain-English spec of what it does, so any engineer can rewrite it in a modern language. Around 280,000 developer hours saved in five months.
A workflow-automation company put AI on onboarding. New hires were productive in around two weeks instead of months.
Every one started with a diagnostic
Four different problems, and none of them started with “let’s use AI for coding.”
Every one of them started with a diagnostic: where does the work actually pile up? For one it was the review queue. For another, the test suite. Then the mountain of legacy code nobody wants to touch, and the first ninety days of a new hire. A different answer each time. Then they put the agent exactly on that spot.
Most teams start at the IDE
Most teams skip that step. They install the agent at the most visible part of the pipeline, usually the IDE, and hope the wins compound. Often they don’t. The wins land where the constraint actually was, and that’s rarely the IDE.
The diagnostic travels, the playbook does not
So the move that travels from these four cases is the diagnostic that comes before the playbooks. Where does your team’s work pile up this quarter, honestly? Answer that first and it picks the playbook for you.
Copying the playbook without the diagnostic is how a team ends up with a fast IDE and the same slow quarter.
What did your team optimise first, and was it actually the constraint?
Source: AI engineering case studies from Faire, Meta, Morgan Stanley and Zapier, 2025 and 2026.
Further reading
Written by
-
Benedikt Stemmildt · Founder & Co-CEO
More on Adoption
- Give every team one AI hypothesis instead of generic training Generic AI training across 300 people produces compliance, not adoption. One written hypothesis per team, tested on a retro cadence, makes the rollout correct itself.
- Run the migration you keep postponing as the training Train agentic engineering on the legacy migration the team has postponed for years, not on a sandbox. The team learns on its real codebase while the debt shrinks.
- When the CEO sits in the developer training AI stopped being a tooling purchase and became an investment in the environment: the harness, the context layer, the trained team, the hiring filter. A CEO doing the exercises next to their staff engineers shows the shift.
Test the constraint in your system
The Factory Readiness Assessment connects this research to your value stream, your data, and the next decision.
Training on this topic: Agentic Engineering for Engineering and Product LeadersWant to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check