AI is speeding up developers; software delivery is not keeping pace
For two years, the pitch for AI coding tools has been simple: developers write code faster, so companies ship faster. Atlassian’s latest global survey suggests the second half of that sentence is not holding up. Individual speed is up. Team-level delivery, in many organisations, is not.
The headline numbers
Across the survey, 87% of developers say AI helps them ship work faster. Yet 72% report the same or more friction than before AI arrived. Both findings can be true at once, and that tension is the story of the report.
The sharpest data point is a year-on-year drop. In 2025, 68% of developers said AI saved them at least 10 hours a week. In 2026, only 25% say so. The tools did not get worse. What appears to have changed is where the saved time goes.
Where the time went
According to the report, 65% of developers now regularly spend time reviewing, debugging and correcting AI-generated code. That is a new category of work that barely existed before these tools became mainstream. Atlassian also points to slow code reviews and merge problems further down the pipeline, which can cancel out gains made earlier in the process.
This is a classic systems problem. Speeding up one stage of a pipeline does not speed up the pipeline if the bottleneck sits elsewhere. Writing code faster simply moves the queue to review, integration and release.
The old problems are still the biggest ones
The top sources of developer friction in 2026 look much like last year’s:
• Finding information: 51%
• Meetings: 50%
• Context switching: 45%
• Technical debt: 44%
• Unclear requirements: 40%
Atlassian notes that AI could help with several of them, from knowledge discovery to mapping dependencies to automating handoffs, but most deployments remain focused on code generation. In other words, companies have pointed AI at the part of the job that was already the most visible, not necessarily the part that was most broken.
“AI can help developers move quicker, but faster code is not the same as better software delivered sooner,” said Andrew Boyagi, Atlassian’s Customer CTO.
Usage is not strategy
Perhaps the most consequential finding concerns leadership. Only 17% of developers and 23% of engineering leaders say their organisation has a clearly communicated and implemented AI strategy.
The measurement gap is just as telling. 60% of developers say AI usage is treated as a productivity signal, but only 42% of leaders believe they have the right metric to separate genuine gains from mere AI activity. When adoption itself becomes the target, teams are rewarded for using the tools, not for the outcomes the tools were meant to improve.
There is an upside. Developers at organisations with a clear AI strategy are twice as likely to report less friction than before AI was introduced. Strategy, not tooling, seems to be the differentiator.
Why it matters for India
Atlassian frames the findings as especially relevant for India, home to one of the world’s largest developer populations and a fast-growing network of Global Capability Centres. Large, distributed teams working on high-volume codebases are exactly the setting where review queues, merge conflicts and information silos can swallow individual productivity gains. For GCCs under pressure to show returns on AI investment, the report is a warning that tool rollout alone will not deliver them.
Those caveats do not erase the pattern, but they suggest treating the 68% to 25% drop as a strong signal rather than a precise measurement.
The takeaway
The report’s core message is that AI has improved a task, not yet a system. Organisations that want real returns will likely need to do three things: aim AI at the whole delivery flow, including information retrieval, requirements and handoffs; measure team outcomes rather than tool usage; and communicate a clear strategy that developers can actually see.
The question for engineering leaders is no longer whether their developers are using AI. It is whether anything is reaching customers sooner because of it.