Nearly all Java developers surveyed by BellSoft are using artificial intelligence in their workflows, but few are willing to give the technology independent authority over code approvals, highlighting a persistent gap between AI adoption and trust.
The Java platform vendor surveyed 360 developers at Devoxx Belgium 2026 and found that 98% use AI in development. However, fewer than 1% of teams allow AI to approve most pull requests, according to the research.
The findings highlight potential opportunities for managed service providers (MSPs) and software development partners to help customers implement AI coding tools while maintaining security, code quality and human oversight.
AI adoption reaches 98%, but code approval remains limited
BellSoft surveyed 360 Java developers at Devoxx Belgium 2026 and found that 98% use AI in their development workflows.
That exceeds the 83%-90% adoption rates reported in the 2026 JetBrains and Stack Overflow developer surveys. It also marks a significant increase over BellSoft’s own 2024 Devoxx survey, in which 74% of respondents reported using AI for code writing.
However, adoption does not necessarily mean developers trust AI to make decisions independently.
| Development activity | AI adoption |
| Code review | 77%, but only 25% let AI review most pull requests first |
| AI-assisted debugging | 93% |
| Use of real debugging tools | 24% |
| AI approval of pull requests | 44% willing |
| AI approval of most pull requests | Less than 1% of teams |
The distinction matters: developers are comfortable using AI for recommendations but remain reluctant to give it operational authority.
Greater AI autonomy brings productivity gains and risks
BellSoft’s findings suggest that developers who give AI more responsibility can see larger efficiency improvements. Among developers who let AI use real debugging tools, 40% reported saving at least half their coding time, compared with 23% among those who did not.
Overall, 61% said AI saves them more than a quarter of their coding time. Still, incorrect code and hallucinations remain the biggest concern, cited by 36% of respondents. Security followed at 27%.
MSPs see opportunities in AI governance and security
For MSPs, MSSPs and software development partners, the findings point to an opportunity beyond selling AI coding assistants. Customers need help establishing the infrastructure and governance that allow AI tools to work safely across development environments.
Potential service opportunities include:
- AI development governance: Establishing approval rules, access controls and human oversight for AI-generated code.
- Secure Java environments: Managing trusted runtimes, libraries and container images to reduce software supply chain exposure.
- AI-assisted debugging: Integrating development tools that let AI investigate real application behavior under controlled permissions.
- Code quality management: Building automated testing and review processes that catch incorrect suggestions before deployment.
One particularly important finding is that developers most concerned about AI security were also less likely to use AI in code review. Eighteen percent of this group had no regular code review process, compared with 7% of less security-concerned developers.
That creates an opportunity for channel partners to address security concerns without prompting customers to abandon AI altogether.
Software supply chain controls become more important
BellSoft CEO Alex Belokrylov argues that AI-generated software can expand the range of open-source libraries and images used by applications, increasing the potential attack surface. His recommendation is to direct AI toward trusted, maintained catalogs of runtimes, images, and libraries.
This is an important consideration for enterprise customers. AI-generated code can introduce dependencies that developers might not otherwise select, making dependency management, vulnerability scanning, and software provenance increasingly important.
BellSoft’s survey also found that 70% of developers expect AI to write most production code by 2030. That expectation gives service providers a reason to start preparing customers for more autonomous development workflows now.
For developers, the immediate benefit is faster coding and debugging, but greater AI autonomy will require stronger validation and oversight. For businesses buying software services, the concern extends to the reliability and security of applications built with AI assistance.
Customers should ask development partners how they validate generated code, manage third-party dependencies, and control AI access to production environments.
The next stage of AI adoption in Java development may not depend on getting developers to use more AI. With adoption already at 98% in BellSoft’s survey, the bigger challenge is helping them trust AI with more responsibility without compromising the security and quality of the software they deliver.


