Your engineers got twice as fast. Your lead time only moved 4%
Bharat Sharma · 11 November 2026
Read the originalAn engineering leader found that AI roughly doubled coding speed on one initiative but cut its lead time by only 4%, because coding was a tenth of the time.
Bharat Sharma is a senior engineering leader with more than 17 years in platform and data engineering at large product companies. Writing in LeadDev, a publication for engineering leaders, Sharma recounts that a peer asked how business outcomes had changed after near-complete adoption of AI coding tools. Sharma had no answer, and then compared an initiative coded mostly with AI against a similar one coded without it.
- Without AI assistance, coding took about 10% of the initiative's total lead time, so a near doubling of coding speed shortened lead time by about 4%. Sharma calls this one comparison in one organisation.
- Surveys point the same way. A 2025 Microsoft study of 484 developers found coding took about 11% of a typical week, and Atlassian's 2025 survey of 3,500 developers found 16%.
- Sharma draws on Eliyahu Goldratt's theory of constraints: speeding up a step that is not the slowest one does not raise what a team delivers. It can instead add pressure on code review, decisions, other teams' backlogs and approval gates.
- Sharma advises reporting lead time, split into waiting, active work and rework, next to AI adoption figures, recording a specific reason against every wait, and tracking rollbacks and incidents.
An engineering manager judging AI coding tools can map where work waits before buying more of them, as Sharma advises. The 4% figure comes from a single comparison in the author's own organisation, which the article does not name.
Have engineering managers report lead time split into waiting time, active work and rework next to AI adoption figures, with a specific reason recorded for every wait.
Before buying more AI coding tools, compare lead time on similar pieces of work from before and after adoption, including tickets still unfinished.
Derived by Working Surface from the article; more in the Playbook. Source line: Before buying more AI tooling, find out where your work waits.
16 November 2026: A team learns what AI is worth by counting what its people kept from its work, by having experts correct what agents read, and by recording where work waits.