There is a question many QA professionals are thinking about, but few are comfortable asking openly:
If AI can generate test cases, write automation scripts and analyse failures, what exactly will testers be needed for?
This is the elephant in the QA room.
Ignoring it will not make it disappear. Giving reassuring answers such as “AI can never replace testers” will not prepare anyone for what is coming either.
The honest answer is more complicated: software will still need quality assurance, but many activities traditionally performed by testers are changing. Some will require less human effort; others will demand stronger judgement and new skills.
Quality remains the end goal. What is changing is how that quality is achieved—and what organisations expect from QA professionals.
The concern is reasonable
A QA engineer spends several hours reading a requirement and preparing test cases. Another person gives the same requirement to an AI assistant and receives a detailed first draft in minutes.
An automation engineer takes a day to create and debug a test script. AI produces a usable first version in a fraction of that time.
AI can already help teams:
- Review requirements
- Draft positive, negative and boundary tests
- Generate test data
- Write Selenium, Playwright and API automation code
- Explain logs and error messages
- Summarise failures, defects and test results
The output may be incomplete, repetitive or based on assumptions. But AI does not have to be perfect to change how work is planned and staffed. If one person can complete more routine work with it, organisations will reconsider the skills and capacity they need.
The question is therefore not only, “Will AI replace testers?” A better question is:
Which parts of my work are becoming easier for AI to perform, and what value will my organisation expect from me next?
A QA role is more than one activity
“Tester” is a role. Writing a test case is an activity.
Executing a regression test is an activity. Preparing a report is an activity. Investigating why customers cannot complete a payment is a quality problem requiring product knowledge, system understanding, evidence and judgement.
AI usually changes individual activities before it changes an entire profession. Those activities will not all be affected at the same speed.
| AI can accelerate | Strong human involvement remains important |
|---|---|
| Drafting test cases | Understanding unclear business intent |
| Generating test data | Finding information missing from requirements |
| Creating routine automation code | Deciding which failures create the greatest risk |
| Writing basic SQL and API requests | Challenging incorrect assumptions |
| Summarising logs and results | Investigating unexpected system behaviour |
| Producing routine documentation | Deciding whether quality evidence is sufficient |
AI is particularly effective when the input is clear, the output follows a familiar pattern and limited business context is required. A well-written login requirement, for example, can quickly be converted into standard scenarios and a basic automation script.
This means routine artefact production is becoming less distinctive. Statements such as these may no longer carry the same value:
- “I can write test cases.”
- “I can create Selenium scripts.”
- “I can execute regression tests.”
- “I can prepare test reports.”
These abilities are not useless. They are becoming the starting point rather than the differentiator.
The more valuable questions are:
- Can you decide whether AI-generated tests are sufficient?
- Can you identify the risks AI missed?
- Can you provide the right business and technical context?
- Can you recognise a convincing but incorrect answer?
- Can you integrate AI into a controlled QA workflow?
- Can you investigate and recover when the solution fails?
This boundary will continue to move. The durable lesson is:
Do not build your entire career around activities that AI can reproduce easily.
Manual testers are not the only people affected
It would be a mistake to think that only manual testing is under pressure.
AI can generate routine code too. An automation engineer whose main value is converting written tests into scripts is also exposed. Knowing Selenium, Playwright or a programming language remains useful, but tool syntax alone offers less protection when AI can produce a reasonable first version.
Manual testers, automation engineers, QA leads and test architects will all face rising expectations. Organisations will move at different speeds, but the direction is clear.
Faster development increases the pressure on QA
AI is not changing testing in isolation. It is also helping developers create and modify software faster.
That can mean:
- More code entering the system
- More frequent changes
- Shorter review cycles
- Less time for QA to understand each change
- More AI-generated code that looks correct but still needs verification
Traditional QA processes may become the new bottleneck. A team cannot respond simply by generating more test cases. It must decide what matters, choose the right coverage and obtain useful quality evidence quickly.
Google’s 2025 DORA research describes successful AI adoption as a systems problem rather than merely a tools problem. Local productivity gains must translate into better product performance instead of creating downstream confusion. For QA, faster coding without a stronger quality system can move the bottleneck rather than remove it. 2025 DORA State of AI-assisted Software Development
This creates an important paradox:
AI may reduce the effort required for individual testing activities while increasing the need for intelligent Quality Engineering.
So, what happens to testers?
Nobody can honestly promise that every QA job is safe or reliably predict how many will disappear.
The most realistic direction is:
Reduced human effort
- Repetitive execution
- Routine documentation
- Basic test generation
- Standard automation implementation
AI-augmented work
- Requirement analysis
- Test design and automation
- Failure investigation
- Test-data and coverage analysis
Growing quality responsibility
- Evaluating AI-generated output
- Testing AI applications and agents
- Assessing non-deterministic behaviour
- Testing security, privacy, performance and cost at scale
- Governing human review and accountability
Software quality will not become unnecessary. But organisations may need fewer people for repetitive work and more people who can deal with greater delivery speed, complex systems and AI-specific risks.
Learn to ride the elephant
AI can write test cases. It can generate automation code, analyse logs and prepare reports. But that does not make quality automatic.
Someone must still:
- Understand the product
- Identify the real risks
- Challenge assumptions
- Evaluate the evidence
- Decide whether the outcome can be trusted
This does not mean the QA role will remain unchanged. The work is changing, and expectations are rising.
Testers must move from producing test artefacts to solving quality problems. Writing more test cases or scripts will not be enough when AI can produce them faster. The greater value lies in knowing what should be tested, why it matters, how AI can help and when AI should not be trusted.
AI readiness also requires more than awareness. Reading about AI, collecting prompts or watching demonstrations is not enough. It requires practice in applying AI, evaluating its output and integrating it responsibly into QA work.
The strongest professional position is:
Strong QA fundamentals + domain knowledge + automation understanding + practical AI capability
In other words:
Quality expertise combined with AI capability offers the strongest position.
If AI is the elephant, competing with its strength is pointless. QA professionals must learn to guide it: give it direction, question its output, control its actions and apply it to problems that matter.
The elephant in the room is not whether AI can test. It is whether today’s testers are preparing for how quality will be engineered tomorrow.
In Part 2, we examine an important distinction: AI can generate tests—but can it assure quality?
Ask yourself
If AI generated your test cases and automation tomorrow, what quality decision would your organisation still depend on you to make?