The New Frontier: Managing Human+AI Agent Software Development Teams

AI has changed the pace and makeup of software development so fast that many companies face huge management challenges as they transition from purely human teams to human + AI teams while maintaining application quality.

Our research shows almost 7 of 10 U.S. software experts say AI writes or accelerates 41% or more of their code. That’s up from 43% of experts who said the same in January. Also, non-technical colleagues are increasingly contributing to codebases using agentic coding tools while autonomous AI agents join teams as active contributors.

These shifts carry major implications for application quality and application integrity, which means applications work as they’re intended. That’s because application quality was once centralized in a focused function composed primarily of humans. A human developer wrote code, understood its dependencies, and knew what else in the system it might touch. The chain of custody started and stopped with a human.

AI Picking Up More Work

Now, AI agents write more of the code and non-technical contributors can make more changes. Using natural language tools, a product manager can describe what they want and make it happen. But they may not understand whether what they changed broke an integration downstream or introduced an edge case nobody tested for.

Also, AI agents are active participants in the development and testing process, sometimes working autonomously toward a goal. As AI and agents absorb more of the work, the human focus shifts toward oversight, judgment, and the higher-order decisions that define what quality means. In many ways, human developers and quality experts will increasingly act less as the creator and tester, and more as the verifier. While who or what does the work changes, a chain of custody for software quality still exists. The entity at the end of that chain must still be human. That’s because ownership requires judgment, context, and the ability to answer for a decision after the fact.

Sharing Responsibility for Quality

This all means that application quality will increasingly become a shared responsibility across hybrid human-AI teams. For executives, managing quality across this new landscape is essential to gaining and keeping a competitive advantage. The imperative is to ensure that application integrity does not erode as AI picks up more of the work.

Before leaders can effectively manage hybrid human-AI resources, they need to address four priorities:

Recognize human concerns. The fact that AI agents are part of the workforce may not be perceived as a good thing by human workers. Some may be skeptical of AI’s abilities. They may have tried AI tools that fell short. With training, education, and a human-in-the loop awareness, those concerns can be alleviated. Recognizing the concerns will be the first step to successfully alleviating them.

Invest in the right tools. The right tools can make software and application quality accessible regardless of whether a team is suited for a fully autonomous solution, a manual one, or a hybrid of both. It falls to leadership to assess which tools fit which team, depending on AI maturity level. A team early in its AI journey might need a hybrid setup, like AI-assisted test generation with manual review gates. A more mature team can push toward greater autonomy. There is no one-size-fits-all tool solution across the software development lifecycle — from creation, to testing, to validating quality.

Double down on quality in testing. It can be exciting to get more applications to market faster given the speed of AI coding. But leaders must pay close attention to the business impact of applications that fail. SmartBear’s 2026 State of Software Quality and Testing surveyed 1,436 U.S. and U.K. leaders and practitioners who use AI in development. It reveals that 46% of teams have shipped AI code that later failed in production. To keep quality on pace with development, autonomous testing tools can fortify human quality teams. Also, upskill quality assurance teams on AI so they can work alongside it and rigorously validate its outputs.

Step up the fundamentals. Poor documentation, unclear direction, and fragmented toolchains will reduce productivity even as individual AI tools improve. AI-empowered teams will never deliver 10x improvement until the proper controls are in place. Be clear on what your processes are today before deciding how — and whether — AI can make a positive difference. Robust governance must underpin everything. These are the foundational structures and policies that ensure AI is used ethically, responsibly, and transparently.

Discerning and Nimble

It can be tempting to deploy AI everywhere and embed it in every team. Whether that’s the right move depends entirely on where each team and company is in its AI journey. Leaders must honestly assess their current state before prescribing an approach.

What’s more, AI will never be a set-it-and-leave-it technology. It evolves too quickly, and it will demand new levels of organizational nimbleness. Leaders must continuously revisit how they manage AI and its role within their organization. The right approach this year may not be the right one next year.