KairosJetBrains Kairos

The Big Handover

96 in 100 surveyed developers use AI at work. Adoption has moved faster than the review, coordination, and governance around it. The challenge lies in how AI-assisted work is checked, controlled, and owned.

September 2026An essay by Kirill Skrygan, CEO of JetBrains7-minute read
what changed?who approved?who has context?what did it cost?
The IDE
The IDE EraThe bulk of the coding loop was concentrated within the IDEWrite. Run. Debug. Refactor. The code and its immediate context were close enough together for one developer to follow.
Keep scrolling

At the developer level, the main handover problem is context switching. For organizations, the issue is how to assess business value. Our State of Developer Ecosystem 2026 Survey examines the strain at both levels.

The NumbersDevelopers report review friction. Organizations report governance gaps
01Where developers report friction
More time reviewing AI output0%
More time correcting or rewriting it0%
Harder to spot errors in what it wrote0%
Technical debt from AI-generated code0%
Less confidence in the codebase0%
Unclear who authored the code that landed0%
Overhead getting access, licenses, and tokens0%
Frictions reported by developers who use AI regularly at work. Half of them spend more than a quarter of the coding time on reviewing AI output. Data taken from the State of Developer Ecosystem 2026 Survey.
02Oversight lags adoption
0%of companies refrain from using autonomous agents due to data privacy and security concerns, rising to 57% among organizations with over 1,000 employees.
0%of organizations do not measure the return of investment in AI at all. Separately, 22% of tech managers at the largest companies say they do not know the ROI.
0%of tech managers who track the impact of AI consider token usage. However, only 30% of them find it insightful.
State of Developer Ecosystem 2026 Survey

Those pressures sit at opposite ends of the same workflow. It starts at a developer’s desk and, as agents take on more, ends beyond it.

96%use AI at work
OrganizationWhere it is answered for
TeamsWhere it is coordinated & automated
DevelopersWhere the work is done
The Baseline96% of employed developers use AI at workThe handover starts at the developer’s desk, because that is where AI entered the workflow.State of Developer Ecosystem 2026 Survey
Keep scrolling

These systems can make delegation visible, controlled, and auditable. They are still unable to take full responsibility for results.

When production breaks, an accountable person is on callThe incident is owned by a role inside the organization — not by the model. That owner needs enough evidence, authority, and control to diagnose, contain, and learn from the failure.
Keep scrolling

Helping developers create great software is still the mission. AI widens it — from individual productivity to trusted, organization-wide production.

Agents will do more of the work. But the handover is only complete when a named owner has the evidence and authority to answer for the result

Kirill Skrygan, CEO of JetBrainsSeptember 2026

Thoughts from the Lab

Quo Vadis, Code Review?By Stephan Lukasczyk, Agent Testing and Evaluation Researcher
Verifying suggestions, refining prompts, and reworking generated code can take up to half of a developer’s time.