A brand new research of 700 engineering practitioners and managers throughout the U.S., U.Ok., France, Germany, and India reveals a basic shift within the software program growth panorama. Whereas generative AI has accelerated code manufacturing, it has launched a large “invisible” workload that conventional productiveness metrics fail to seize.
For many years, technological shifts just like the web, the cloud, and DevOps modified how software program was distributed and deployed, however the core cognitive act of growth remained largely the identical. Generative AI has damaged this sample, shifting the transformation to the cognitive layer. Builders have shifted from being the first authors of code to changing into validators of machine-generated output.
Based on the 2026 State of Engineering Excellence report from Harness, 31% of a developer’s day is now consumed by AI-related invisible work. This consists of deeper scrutiny of code high quality, elevated accountability for downstream outcomes, and complicated judgment calls relating to when to belief or override AI. Regardless of this, established frameworks like DORA metrics and cycle time weren’t designed to measure these new necessities.
The information highlights a big “productiveness offset.” Whereas AI improves gross output quantity and shortens cycle instances, 81% of engineering leaders report that code evaluate time—typically considered as overhead or “toil”—has risen sharply since deploying AI. This rise in validation effort typically exists exterior the measurement course of, resulting in systemic friction.
Builders recognized the highest sources of this AI-driven friction as reviewing AI code for accuracy (53%), fixing delicate bugs (52%), and explaining AI-generated code to teammates (48%). Mockingly, solely 38% of organizations truly monitor the time spent reviewing AI-generated code.
There may be additionally a stark disconnect between management and practitioners. Whereas 94% of respondents agree that tech debt, validation time, and burnout are lacking from present metrics, managers typically report extra favorable circumstances than these doing the work. Moreover, 54% of builders worry that AI productiveness information will likely be used towards them in particular person efficiency evaluations.
To bridge this hole, the report suggests 5 key beginning factors for organizations in 2026:
- Measure validation work: Monitor debugging overhead and context-switching alongside output.
- Prioritize ship fee: Distinguish between producing code quantity and transport precise worth.
- Audit frameworks: Deal with excessive confidence in incomplete measurement methods as a danger sign.
- Plan for complexity: Anticipate elevated wants for governance and safety evaluations as AI scales.
- Construct belief: Set up clear coverage guardrails about information utilization to encourage developer partnership.
As AI instruments eat a bigger share of engineering budgets, the business should evolve its productiveness frameworks to account for the true shift in effort.
