Working Surfacelink · 18 Nov 2026
02
Two barges side by side on a canal, one pouring twice the smoke from its funnel yet level with the other at a marker post, while a silhouetted figure on the towpath measures the gap with a knotted rope, the whole scene drawn with wide empty margins on every side.
LeadDev

Stop blaming AI tools for flat velocity

Arjun Garg · 17 November 2026

Company story · decisions · agents

Read the original
Takeaway

Arjun Garg, a senior engineering manager at Intuit, found that the heaviest AI spenders did not merge more pull requests, and proposes a test to learn why.

Summary

Arjun Garg, a senior engineering manager at the financial software company Intuit, wrote in LeadDev about delivery that stayed flat after AI tools were adopted. Garg compared spending on AI credits with merged pull requests and ticket closure rates, expecting the heaviest spenders to lead. They spent about twice as much as a middle group but did not merge more pull requests.

Key points
  • The heaviest-spending group spent roughly twice what the middle group did, and the middle group was slightly ahead on merged pull requests.
  • Garg notes that a count of pull requests measures volume and not value, so heavy users may be doing harder work.
  • Garg separates two causes. In a capability gap, the tool cannot do the work even with good context and review; in a habit gap, the team falls back on old ways of working.
  • The proposed test gives an AI agent the next hard but bounded piece of work, with a person setting the design, constraints, acceptance criteria and review standard.
Implication

An engineering leader asked whether AI spending pays can run Garg's test on one piece of roadmap work before pricing it as senior engineers' time. Garg reports the internal data loosely and without a full method.

Plays
Engineering

Before an engineering leader prices a hard but bounded task as senior engineers' time, have an AI agent attempt it first, with a person setting the design and review standard.

Engineering

Read AI credit spending next to merged pull requests and ticket closure rates, and do not treat spending alone as a measure of delivery.

Derived by Working Surface from the article; more in the Playbook. Source line: Spend can tell you that AI is being used, but it cannot tell you why velocity has or has not changed.

Source issue

18 November 2026: when AI agents make a team's output steady, the team's attention moves to the places where work stalls and to the people who check it.