A survey of 730 developers: 75% have tried AI, 35.6% are AI optimists, 38.4% are AI pessimists, freelancers use AI more, 50%+ of employers pay for AI, and more
We surveyed 730 coders and developers about how (and how often) they use AI chatbots on the job. The results amazed and disturbed us. Bluesky: @smcgrath.phd Bluesky: Scott McGrath / @smcgrath.phd : In a Wired survey of 730 software engineers: — • ~75% have tried AI (17% use it all the time) — • ~75% of entry-level coders say they are AI optimists — • ~50% of mid-career devs were AI pessimists — • Only 4% use AI without their employers' knowledge …
Context & Ripple Effects
This survey captures developer AI use moving from experimentation toward workplace-sanctioned tooling: roughly three-quarters of respondents had tried it, while more than half of employers were paying for it. That extends the earlier finding that 70% of developers used or planned to use coding AI, even as trust was limited in the 2023 developer survey on coding AI.
The split between optimistic entry-level developers and pessimistic mid-career respondents mirrors broader workplace unease: a survey of US workers found worry about AI’s job effects outweighed hope. Related developer coverage also indicates that adoption can rise without a corresponding increase in confidence in output accuracy.
First-order effects
- Employers paying for AI tools shift usage from informal individual experimentation to approved, budgeted developer tooling; only 4% reporting undisclosed use suggests governance is already part of deployment.
- Freelancers’ higher reported use gives independent developers more immediate exposure to AI-assisted workflows, while sentiment differs sharply by career stage among employed developers.
Second-order effects
- Tool vendors must compete not only on developer appeal but on employer requirements such as managed access, procurement fit, and confidence in output quality.
- Teams will need to reconcile uneven adoption and skepticism across seniority levels, making review practices and workflow integration more consequential than raw access to chatbots.
Third-order effects
- If employer-funded adoption continues, AI assistance is likely to become a standard layer of the software-development workspace rather than a shadow productivity tool, with value concentrating in products embedded in existing workflows.
- The persistent gap between use and confidence may make verification and accountability the durable constraints on AI’s role in engineering, rather than developer awareness or availability.
The trend: Developer AI is transitioning from optional experimentation to employer-governed, workflow-native infrastructure, while trust and workforce sentiment remain unresolved.