AI is not removing testing. It is changing who does it, when it happens, and what QA leadership is for.
This is Part 5 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 hardest part of this transition may not be technical.
Directors of QA are often measured through the visible machinery of the old system: team size, test counts, automation percentage, defects found, execution time, escaped bugs, and release status.
Moving toward confidence engineering can make the organization look smaller or less busy before the business understands that it is becoming more valuable.
The Director must explain the transition without defending activity for its own sake.
Do not argue that the company needs twenty people because it historically had twenty people. Explain which risks require judgment, which evidence must remain independent, which capabilities the business cannot afford to lose, and how AI lets the team cover more behavior at lower cost.
Do not celebrate the number of tests generated. Show which decisions became faster, which incidents were prevented, which risky releases were contained, and which customer failures became permanent evidence.
Do not promise that AI eliminates uncertainty. Show that the organization can measure uncertainty and make better decisions with it.
Watch the Meetings
When engineering and product leaders stop inviting QA into conversations about architecture, coding agents, release strategy, or AI-enabled development, that is not a harmless calendar change.
It means they no longer expect QA to shape the solution.
In several situations I have heard about, exclusion from those meetings was the clearest early signal. The rest of the organization spent the next few months building a plan without QA. By the time the Director recognized the problem, the company had already decided where the new work would live.
The role was gone within roughly a quarter.
Engineering and product teams do not always confront QA immediately. They often build an alternative first. They create their own loops using Claude, Codex, and browser agents. They gather enough evidence to work without the old process. Then they decide which parts of the QA organization still have value.
The longer QA waits to join that work, the less influence it has over the result.
The Next-Job Problem
A QA director who leaves today will be asked in the next interview, “How did you transform quality for AI?”
If the answer is that they preserved the old process and waited for the tools to mature, the next role becomes much harder to obtain. They lost both the position and the best available opportunity to build the experience the market now expects.
My direct messages are full of leaders living in this “in-between land.” Privately, they describe fear, resistance, stalled experiments, team conflict, and uncertainty about their own role.
Publicly, their LinkedIn posts and conference talks say they are embracing AI, transforming quality, and moving quickly.
Sometimes the gap is hidden inside an evaluation project. A leader creates a three-month process to compare every AI testing vendor, many of which are already obsolete, shallow wrappers around older approaches, or mostly marketing.
The process looks responsible. In practice, some of these projects are less about learning than delaying accountability and buying breathing room.
The tools change before a slow bake-off ends. The organization spends another quarter evaluating yesterday’s options while engineering and product learn by doing.
Eventually, the company stops needing permission to work around QA.
The New Mandate
The title does not need to change immediately. The mandate does.
The Director should own the discipline that turns AI-generated activity into justified business confidence.
That includes:
quality strategy
evidence architecture
risk models and release criteria
human-review design
independent validation
production feedback loops
development of people who can operate across product, engineering, data, and testing
Some current work should be automated. Some should be stopped. Some people will need to learn new skills. Some will discover that the judgment they developed through years of testing is suddenly more valuable than the scripts that once defined their jobs.
The Director cannot protect the team from this transition by pretending it is not happening. The Director can lead the team through it by making the destination clear.
Nobody has a finished playbook. Engineering, product, and executive teams are experimenting too. This is a rare period when leaders are expected to try things, learn, and change direction. Failure is acceptable when it is fast, visible, and produces evidence.
The dangerous move is waiting.
AI can generate more code, more tests, more outputs, and more analysis than any organization can consume.
The scarce thing is confidence.
The Director of QA who learns to engineer that confidence does not become less relevant. They become one of the most important leaders in the company.
That is the larger arc of this series. The QA leader begins trapped between preserving the past and surrendering to automation. The path out is to build the people, measurement systems, independent evidence, and leadership mandate that let the whole company move faster without pretending risk disappeared.
Learn more in the Testing AI knowledge guide.
Lead the Transformation Before Someone Else Does
At IcebergQA, we help engineering and QA directors establish the mandate, evidence, and early results needed to lead AI-first quality. Our experience spans Microsoft, Google, startups, testing at massive scale, building AI testing systems, and testing the AI systems that now generate software.
Our 30-day AI quality transformation sprint aligns QA, engineering, product, and executive leadership around a practical confidence strategy. We assess the current organization, identify immediate AI opportunities, launch working quality loops, define the new leadership measures, and give the Director a credible transformation story backed by evidence.
IcebergQA is here for leaders ready to move now. Engineering and product are not going to wait.
—Jason Arbon, IcebergQA