The introduction of AI agents into the development process has radically changed how engineering teams operate — from the roles of developers and team leads to QA structures and the responsibilities of engineering leadership. Productivity has indeed grown by approximately 2x, but this shift introduced a range of new risks: burnout, increased cognitive load, and the need to rethink roles and principles of team organisation. In this article I describe how these changes affect each role in the development process.
Developers
It might seem that the work of developers should speed up dramatically with AI agents. In reality, acceleration exists, but not nearly as dramatically as expected. Here’s why.
Let’s consider the development workflow before and after AI agents.
Before AI agents, a developer would first think through the task and then perform monotonous coding. Now, the developer is engaged in high-load cognitive work at every step. Deep thinking, prompting, and review all require full concentration and continuous decision-making. 30–40% high load before vs 70–80% now. Developers now make decisions constantly — and that’s the core problem.
Cognitive load and decision fatigue
The Strength Model of Self-Control shows that “all acts of controlled thinking draw from the same limited cognitive resource, and once it is depleted, complex reasoning and error detection degrade sharply.” AI removes the routine part (typing code) but leaves the hardest part: thinking, checking, deciding.
As a result, the developer needs more time to rest. Otherwise attention drops and bugs slip in after the agent’s coding pass. So if you see a developer frequently in the office kitchen drinking coffee, it’s not because they work poorly — they need more rest after high-load work. Paradoxically: we work faster, but need more rest.
Context switching gets worse
The same research demonstrates that task-switching and maintaining multiple goals accelerate this depletion, leading to noticeably worse performance. And when a developer executes tasks faster, it means more frequent task switching. In our company we monitor the number of parallel tasks from different epics per engineer and aim not to exceed two — significantly reducing context switching.
Quality Assurance
The SDET role is disappearing
More companies realise that developers or QA engineers can now produce automated tests quickly with AI assistance. QA writes test cases, the agent generates automated tests, and developers help in complex scenarios.
Manual testing also decreases
Why manually check a scenario if a single prompt produces an automated test that can be reproduced infinitely? This is not laziness — it’s optimisation. Velocity increases while manual test runs decrease.
There is some homework, though. For example, we load Swagger specifications directly into the agent’s context, allowing it to understand the API structure and generate accurate HTTP requests. You can also generate documentation related to the code, which is then uploaded into the autotest prompt context.
I am sure that in the near future we will have high-quality automation across the entire flow. Agents will generate test cases from requirements, automate them, and run them. QA specialists will act as orchestrators.
Team Leads
Team leads no longer need to write highly detailed specifications — they need developers who can deliver without rigid instructions, because writing a detailed task is already 80% of the work. It’s essentially the same as prompting an agent. A team lead should instead continue providing architectural oversight; without it the agents’ solutions eventually become unsupportable.
AI gives team leads more free time, and they must choose how to use it — hands-on development, research, or deeper product involvement. Each team lead will choose their own path, but I personally prefer leads who focus on research and product direction — this creates long-term impact.
Juniors
For the first time in many years we’re facing a dilemma — do we still need junior developers in the age of AI?
The answer is yes — but only those with high motivation and potential. Many senior engineers lose motivation after years in the industry; not all, but many. And a senior engineer who still has strong energy and drive is rare and extremely valuable. To address this scarcity, we should hire young specialists. A motivated, energetic junior is a long-term investment.
Technical Direction
There are no major changes in our responsibilities. What has changed is that we need to move faster. As always, we must adopt new technologies — but now we have to do it more quickly. We need to hire the right engineers, support them, and guide their growth.
Originally published on Medium on December 10, 2025.