AI is not removing testing. It is changing who does it, when it happens, and what QA leadership is for.
This is Part 3 of The Director of QA Dilemma, a five-part series about one larger transition: the Director of QA must move from defending testing activity to engineering the confidence that lets the company ship AI-generated software.
The Director is squeezed by three forces at once: a team that wants safety, management that wants transformation, and a role that now demands unfamiliar skills.
That is the triple bind.
The Team Wants Safety
Existing team members are reasonably worried that AI may take some of their work or their jobs. People also dislike disruption.
Many testers say they want more freedom, creativity, and learning, but some still prefer the predictability of a regression queue, a known test plan, and familiar release rituals when they face a real choice. Repetition can feel safe. Ambiguity does not.
That does not make them bad people or bad testers. It does mean the Director cannot transform the function merely by announcing access to an AI tool.
The team needs a clear destination, honest expectations, room to experiment, and accountability for changing how it works.
Management Wants Speed
Engineering and product teams are already experimenting rapidly with AI. They accept failed experiments because everyone is learning.
When QA insists that the existing process remain unchanged until the new approach is proven perfectly, QA does not look careful. It looks slow.
The cruel part is that QA leaders often built their reputations by preventing disruption. They kept releases calm, reduced surprises, standardized work, and avoided failure.
Now the organization needs them to experiment, learn in public, and fail quickly enough to find what works. The behavior that made them successful can make them immobile.
The Role Demands a New Identity
Many QA directors became leaders because they were excellent testers or excellent managers. Some deliberately avoided the development path, product-management accountability, or deep technical work.
AI is merging those boundaries.
Confidence engineering requires product judgment, engineering awareness, data interpretation, risk ownership, communication, and the ability to direct coding agents. The Director may need to learn from someone junior who has spent months seriously experimenting with AI.
That can be uncomfortable. A confident leader who believes they have already been around the block may dismiss the person who understands the new terrain best.
When some QA directors are removed, companies do not hire another traditional QA director. They promote the engineer or tester most open to AI and most willing to experiment. That person may not have perfect results. Management would rather work with someone learning forward than someone defending backward.
Stop Preserving Categories
Routine test execution, basic test-case writing, straightforward automation, result transcription, and first-pass failure triage are becoming cheaper.
Much of this work consumed the time of talented testers who could have been studying the product, investigating risk, challenging assumptions, and helping the company make better decisions.
AI exposes an uncomfortable truth: not everyone performing manual testing developed deep exploratory judgment, and not everyone writing automation developed strong product judgment. Job titles hide enormous differences in capability.
Strong manual testers may thrive because they understand ambiguity, user behavior, failure consequences, and the difference between a technically valid output and a good product outcome.
Strong automation engineers may thrive because they understand reliable execution, environments, observability, data, scale, and reproducible evidence.
The most valuable people will cross those boundaries. They will use coding agents to understand code, generate experiments, build harnesses, inspect traces, and investigate failures. They will know enough about the product and business to recognize when the measurement system rewards the wrong behavior.
The Director’s job is not to preserve the old categories. It is to identify who can grow into the new work and give them room to do it.
The Emotional Sunk Cost
A company may have invested millions of dollars and years of effort in automation scripts, manual regressions, environments, dashboards, and processes that no longer make economic sense.
Retiring some of those assets does not mean the original investment was fraudulent or foolish. The technology changed.
The harder sunk cost is emotional. Leaders built careers, reputations, and teams around those systems. Deprecating them can feel like admitting that their life’s work was wrong.
It was not wrong. It solved the problem the company had at the time.
Leadership now means recognizing when yesterday’s successful solution has become today’s constraint.
Do not protect people by promising their work will never change. Protect their future by helping them develop the judgment, technical fluency, and adaptability the new organization needs.
Part 4, Build the Confidence System, moves from people to architecture. It explains what replaces the old test phase when evidence must surround AI-assisted development.
Learn more in the Testing AI knowledge guide.
Transform the Team Without Losing Its Best Judgment
At IcebergQA, we have lived through quality transformations at Microsoft, Google, and AI startups. We understand both sides of this transition: testing complex systems at enormous scale and using AI to replace, accelerate, or radically reshape the work.
Our 30-day AI quality transformation sprint helps leaders assess roles and capabilities, identify the people who can accelerate the change, preserve valuable product knowledge, and create a practical skills and operating plan. The result is not indiscriminate cost cutting. It is a smaller, faster, more capable confidence-engineering function with a clear place in the company’s future.
—Jason Arbon, IcebergQA