Organisations have been adopting AI Coding tools since GitHub Copilot, with many transitioning to Cursor, Claude Code and other tools chasing the best tool for job. Adoption of AI Coding tools has climbed, and spending has climbed faster, but productivity gain is 10%; very different to the 10x gain we might expect from press releases, blog articles and weekend experiences.
The explanation for this difference between expectation and enterprise reality flows from the Theory of Constraints. At any point in time every goal-orientated system has exactly one constraint, and the system’s throughput is the throughput of that constraint. AI will only increase your throughput if it improves your constraint.
This site is a companion to the talk given at Web Direction’s AI Engineer conference, Melbourne June 2026.