AI is not removing testing. It is changing who does it, when it happens, and what QA leadership is for.
This is Part 2 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.
Let’s be blunt: the AI transition should be a layup for an experienced, creative, and intelligent QA director.
Producing code is no longer the hardest part of software. For a growing share of product development, coding is fast, cheap, and increasingly handled by agents. Writing feature lists, button specifications, user stories, and initial product requirements is becoming cheap too.
The old product triad was product management, development, and testing. Product defined what to build. Development built it. Testing decided whether it worked.
Those roles are converging. When AI can propose the product, write the code, create variations, and run much of the validation, the scarce capability is the judgment required to balance speed, risk, customer experience, business value, and confidence.
That sounds much closer to the work of a great QA director than the work of a pure project manager or a pure developer.
Quality used to look slow because coding was the visible work and validation followed behind it. Now AI can generate implementations and product variations almost instantly, while establishing confidence still takes evidence, context, experimentation, and judgment.
Quality is not merely the slow part anymore.
Quality is the hard thing now.
The Wartime QA Director
Ben Horowitz’s The Hard Thing About Hard Things describes leadership when there are no easy answers and distinguishes between peacetime and wartime leadership. AI has created a wartime moment for QA.
That does not mean panic, recklessness, or treating people badly. It means moving with urgency, making difficult decisions directly, leading from the front, and refusing to use process as shelter from accountability.
The details vary by company, but the immediate playbook is short.
1. Join Engineering’s and Product’s AI Transformation
Understand exactly how developers and product managers use AI, which coding agents they use, what workflows are changing, and where they still lack confidence.
Put your team’s judgment, evals, risks, and evidence requirements inside those agents and processes. If QA does not become part of the new workflow, there may be no chair left when the old workflow stops.
2. Build a Team That Accelerates the Crossing
Evaluate who will not merely tolerate the transition, but actively accelerate it.
Plan against substantial downward pressure on traditional QA teams and vendors, even though the number will differ by company. Your leaders need to be AI-aware, AI-competent, visibly experimenting, and willing to take intelligent risks.
Retrain and reposition people who can move. Make difficult decisions about people who refuse. If you do not reshape the team proactively, someone else will reshape it and include you in the change.
3. Turn Existing Assets Into AI Context
Your test cases, automation, plans, bug history, reports, and domain knowledge are not all sunk costs.
Export them from closed tools into open, AI-readable formats. Put them into an approved enterprise AI workspace with the right privacy and access controls. Add the relevant code, product strategy, and business context.
Then ask the agent which assets remain valuable, which have become liabilities, what should be deprecated, and how to communicate the transition. Ask it to produce the plan, schedule, risks, and first experiments.
4. Begin Every Task With AI
Before writing an email, planning a project, reviewing a release, designing a test, or entering a meeting, ask an agent what you should consider given the persistent context.
Ask how the task affects the broader transformation. Use the agent to challenge your message, expose missing risks, and suggest the next action.
This is not blind delegation. It is a new reflex: let AI help you think before you spend human time executing.
5. Lead From the Front
There is no room for a pure manager who delegates all AI experimentation.
Delegate execution, create space for the team, and communicate the roadmap, but remain visibly involved in the tools and the work. I have seen proactive leads stop waiting, work directly with developers, and become the obvious choice to run the new quality function.
That person can take the Director’s job. The Director should be leading that charge first.
6. Make Every Legacy Abstraction Prove Its Value
Test-case management systems, bug trackers, browser-driving frameworks, device harnesses, page-object models, dashboards, and proprietary workflows should not survive merely because the company already paid for them.
Export the data. Ask whether an AI agent can perform the underlying job directly, faster, and with less maintenance. Keep systems that still provide necessary governance, access control, auditability, deterministic execution, or regulatory records. Remove the ones that now create more friction than value.
The AI-first replacement for an old testing framework may not be a newer framework. It may be no framework at all.
Give an approved browser-capable agent a website and ask it to investigate it. Within seconds, it can navigate, observe, form hypotheses, and find issues. The result will not be perfect, but neither is a framework that takes days to configure, weeks to expand, and years to maintain.
The technology is capable enough to begin. Waiting for certainty is itself the decision most likely to make the QA organization irrelevant.
Part 3, Stop Defending the Org Chart, examines the human side of this transition: team fear, uneven capability, role convergence, and the emotional cost of retiring yesterday’s successful systems.
Learn more in the Testing AI knowledge guide.
Put the Playbook Into Practice
IcebergQA helps QA leaders turn urgency into working systems. We bring lessons from Microsoft and Google-scale quality, AI testing startups, and years of applying AI to testing itself.
During a 30-day AI quality transformation sprint, we work alongside your team to connect QA with engineering’s coding agents, convert existing testing assets into useful AI context, identify work that should be retired or accelerated, and put the first AI-led testing loops into production. Your team learns by doing, and leadership sees measurable movement within a month.
Jason Arbon, IcebergQA