How generative AI is eliminating development friction through intelligent automation

Software teams have more automation than ever, but friction still slows down delivery. Routine handoffs, fragmented tools, manual investigation, and unclear quality signals interrupt momentum and make it harder to stay focused on the work that actually matters. In our discussion with QA.tech’s CAIO, Vilhelm von Ehrenheim, he emphasized the same pattern: AI’s real value is in its ability to absorb the work that repeatedly breaks flow.

That flow is now the scarcest resource in development. Every unnecessary interruption forces teams to rebuild context before they can continue. Intelligent automation reduces these breaks by taking on boilerplate tasks, generating first-pass analyses, and surfacing issues earlier, letting developers spend more time reasoning and less time recovering.

Why friction matters

Friction rarely shows up as a single major blocker. It accumulates through tool-hopping, inconsistent signals, and interruptions that reset a developer’s mental model. Research shows that after each interruption, developers take 23 minutes and 15 seconds to regain full focus, and a single unplanned context switch can drain up to 20% of cognitive capacity.

Developers feel the cost immediately. As Jesse Dailey, Solutions Leader at People.ai, puts it:

“AI can handle all of these tasks already. Why waste your time.”

The work he’s describing (documentation cleanup, note-taking, mechanical edits) is exactly the kind of drag that compounds across a week. And it is becoming clearer that friction slows delivery, weakens feedback loops and makes it harder for teams to make confident decisions.

Where generative AI removes friction

Generative AI reduces friction by absorbing the tasks that used to fragment developers’ day. Coding itself represents only 16% of a developer’s time in many organizations. The rest goes to clarifying requirements, searching for information, switching tools, and reconstructing context. This is where the biggest efficiency gains emerge.

Caitlin Colgrove, Co-founder and CTO at Hex, describes one of the clearest examples:

“I want to use data to inform my decisions, but like many organizations, Hex's data is complex, nuanced, and can be challenging to work with if you're not deeply familiar with it.”

AI now handles much of that overhead (query formulation, joining data sources, validating assumptions) making insights accessible without the usual friction.

Jason Sears, CTO at City Lifestyle, captures the shift on the engineering side:

“Being the designer and architect delivering value quickly far outweighs the need for me to have ownership of every line of code.”

The engineering expertise stays intact. What shrinks is the effort required to support it.

Automation enables higher-leverage engineering

Automation is altering what engineers no longer have to do. As workflows compress, developers spend less energy on mechanical steps and more on intent, behavior, and structure.

Both the practitioner quotes and research support this. Half of developers report losing 10 hours per week to tool inefficiencies and information retrieval problems. That reclaimed time shifts the role from typing to directing.

Jason Sears expresses this shift directly, and Nikolai Balba, Founder & CTO at Aquiva Labs, clarifies what it means in practice:

“Developers are evolving beyond just writing code. We are now orchestrating systems where AI agents and humans collaborate to build, operate, and enhance applications.”

Engineers benefit when automation absorbs the routine steps, allowing their time to go toward design, reasoning, and verification.

Automation reduces the cost of quality

Quality used to be expensive, long cycles, manual reviews, brittle tests, late discovery. AI changes this equation by reducing the work required to catch issues early.

But the picture is not one-dimensional. Nearly 45-64% of developers say reviewing AI-generated code takes longer than reviewing human-written code. At the same time, organizations using continuous testing and shift-left practices see up to a 50% reduction in defects. This is the tension many teams are navigating.

Vivek Patel, Co-founder and CTO at Check, notes:

“I both strongly agree with this and am unsatisfied with the state of tooling in our industry to enable high quality code review, more important than ever with AI-generated code.”

And Patrick Glennon, CTO at IDIQ, highlights why performance gains don’t always appear immediately:

“People may not be experiencing the expected boost in velocity… reviewing code generated by the tools offsets productivity gains.”

The cost of quality drops only when review, validation, and automated checks improve alongside code generation.

AI enables true autonomy in development workflows

The next stage of AI in software development centers on coordinated execution across tools and tasks. Agentic systems can interpret intent, plan tasks, and collaborate across tools without constant human input.

Nishant Singh, CTO at DOT, describes this evolution:

“Agentic AI is rewriting that playbook, creating systems that can understand intent, plan tasks, collaborate with other agents, and adapt continuously without constant human input.”

But autonomy brings risk. Studies show that about 30% of AI-generated code samples contain security weaknesses across a wide range of CWE categories.

Aaron Podolny, Co-founder and CTO at Scribe, highlights why this matters:

“Unlike traditional software automation, where business logic is written in code that humans can validate, here an LLM guesses the logic, hiding it from view.”

Autonomy is useful only when systems remain predictable, observable, and continuously verifiable.

Governance, safety, and the new engineering discipline

As automation takes on more of the execution layer, governance becomes a practical necessity rather than a theoretical concern. Teams need clearly defined expectations, boundaries, and validation loops to ensure automated systems behave consistently.

Duri Chitayat, CTO at CINC Systems, explains the challenge:

“Deterministic, machine-driven automation doesn't play nice with socio-technical systems. It struggles with variation, nuance, and the complex, judgment-heavy decisions these systems demand.”

Quality automation helps close this gap. Research shows that teams using continuous AI-based review are 81% more likely to report code quality improvements, compared to 55% of those without it.

This emerging discipline looks less like traditional QA and more like specifying intent and verifying behavior continuously.

The new operating model for software development

Development is shifting toward workflows that move continuously rather than through staged handoffs. As AI takes on multi-step execution, teams spend less time pushing work forward and more time validating direction, interpreting signals, and adjusting in motion. Organizations adopting continuous testing report large operational gains—including 60–70% reductions in testing cycle time—as automated checks keep pace with rapid iteration.

This shift doesn’t eliminate engineering judgment. It changes where that judgment is applied. Instead of managing queues of tasks, teams guide automated systems, verify behavior, and make decisions at the points where expertise matters most. As Patrick Glennon, CTO at IDIQ, noted, performance gains appear when review practices and quality guardrails evolve alongside AI adoption.

Autonomous workflows only succeed when systems remain observable and trustworthy. AI can execute work across the stack, but the value depends on developers knowing whether it behaved as intended. That is why continuous verification becomes central in this operating model. As AI accelerates the rate of change, testing and validation need to match the same pace.

AI QA plays a practical role here. QA.tech’s agent runs alongside development, exploring the application, checking behavior, and catching regressions as soon as they appear. It treats testing as a continuous process rather than a late-stage gate. In a workflow where agents generate and modify code rapidly, QA.tech provides the independent validation layer that ensures those changes behave correctly across real user flows.

‍