AI in Testing is Here — But Is Your Test Automation Stuck in Someone Else's Platform?
BrowserStack's State of AI Testing 2026 report shows 93% of companies using AI in testing workflows. Here's what the data means for QA engineers and vendor lock-in.
Recently I attended the QA Leadership Summit 2026, Leading the AI-Native Transformation, a vendor session hosted by BrowserStack focused on the future of AI in testing. I went in with healthy skepticism and patience knowing I'd be a captive audience for commercials for their AI solutions, but hopeful I'd still come away with vendor-agnostic information as well as knowledge of what BrowserStack is up to in that space. One of the valuable pieces of information they shared was their State of AI Testing 2026 report they compiled by surveying engineering teams on how AI is actually being used today in their testing workflows and where they feel it is going tomorrow. We'll dig into it here. In addition they demoed their AI testing suite which seemed cohesive, but one question nobody in the room asked kept nagging at me: what happens to all that AI-generated test automation if you ever leave BrowserStack?
The State of AI Testing in 2026 — By the Numbers
Their survey underscored AI adoption in testing is no longer on the horizon. It's here. 61% of the companies they surveyed already are using AI in the majority of their testing workflows. 32% of companies are using it for select testing uses. 5% of surveyed companies were curious or still in an exploration phase of AI testing adoption while only 1% responded that they had no plans to use AI for testing at all. That is a rapid uptake from just the prior year.
Surveyed Company AI Usage
So, by the numbers, 93% of companies are actively using AI in many or select workflows in 2026. Based on personal experience, I can see why. When paired with a skilled test engineer it can multiply productivity significantly allowing the tester to focus on less repetitive work and more creative work. This is a similar shift going from manual testing to automated. Or going from doing math by hand to using a calculator.
Where Teams Are Actually Using AI Today
Of the surveyed teams BrowserStack found AI was most used for repetitive tasks. I find this relatable — I often offload repetitive, otherwise time-consuming tasks, freeing me up to either get more done or more creative work done.
| Percent | Task |
|---|---|
| 62% | Generating test cases |
| 58% | Generating test data |
| 57% | Maintaining tests |
| 51% | Authoring tests |
| 49% | Optimizing test suites |
| 48% | Optimizing performance tests |
| 43% | Predictive analytics |
| 42% | Visual testing |
| 41% | Finding and predicting defects |
| 35% | Accessibility testing |
| 35% | Determining test priority / scheduling |
| 32% | Cross browser or mobile app testing |
| 31% | Analyzing test results and automated reporting |
I'm surprised accessibility testing was only 35%. I suspect that might be because companies without public facing websites tend to have less immediate legal reason to invest in A11y testing and others tend to deprioritize it relative other testing until it becomes a must have requirement for a customer, differentiator over a competitor, or a legal concern arises. Regardless, it makes your site more useable for everyone and makes it more testable so I feel strongly it should be a standard and there are already so many tools to catch low hanging issues. That being said, in my own experience, coupling AI with Playwright's @axe-core/playwright assertions allowed me to build out some pretty amazing AI workflows — similar in spirit to getting started with accessibility testing, but now driven by AI rather than manual setup. When combined with the context from the Deque website and the ADO MCP skill I was able to create tickets, with screenshots, impact, and remediation information for developers that was on par with some of the results we received from paid accessibility audits by LevelAccess.
What AI Means for the QA Role
BrowserStack surveyed companies to see how they think AI will change the role of quality assurance engineers. The findings are below:
| Percent | Evolution |
|---|---|
| 50% | Increased need to upskill around AI and AI-tooling |
| 46% | More focus on test strategy, planning, and oversight versus hands on test writing |
| 44% | Increased collaboration with development and AI/ML teams |
| 42% | Shift toward data analysis and interpreting AI output |
| 36% | Role will become more quality coaching and advisement |
| 27% | Less need for testing roles |
To me, this points to fewer QA positions overall. When the role shifts toward strategy, coaching, and AI oversight rather than hands-on test writing, you simply need fewer people to do it — one advisor can serve multiple teams in a way that one test author cannot. The positions that remain will be filled by those with strong QA fundamentals, deep familiarity with AI tooling, and the communication skills to operate at that advisory level.
There have been some interesting trends before AI became ubiquitous that may temper or shape this. Early in my career there was a higher emphasis on testing methodologies and fundamentals. When open-source automated testing tools came along it seemed like fundamentals dropped off in favor of knowing how to write tests in Selenium, Playwright, Cypress, Postman, etc., but those same people never learned how to find a bug, pick apart a requirement, or write a good test case — just write tests of dubious quality. Now, AI can do that, so skill gaps in fundamentals will become apparent very quickly.
On a more positive note, the other recent trend, shift-left, has already reduced tester counts (in general), but the shift-left benefits aren't always actualized, quality suffers, and there are fewer core testers left to do the same or greater amount of work. If the testing is already understaffed, the productivity gains from AI may help make up the deficit by allowing the existing testers to be more productive.
What a Fully AI-Integrated Testing Workflow Looks Like
During the keynote they used the opportunity to show the audience their suite of AI tools in their ideal AI automated testing workflow.
The demo consisted of:
- A hypothetical bug ticket ingested by BrowserStack
- Their AI generated plain language test cases in their test case management system product
- Their AI then turned those into automated tests
- The tests were run
- The results were summarized in their dashboard
The Good
I saw an earlier demo where the AI just stopped at the plain text test case generation. You still had to write the automated test cases. This was cohesive and end-to-end.
Healthy Skepticism
This was a highly curated test site so it's really hard to say how this would work in a real application. Further, there were no repeat runs conducted so for all we know it could be unreliable or edited. There were no speed comparisons done on their AI solution versus a well-written Playwright test for example. Their automation stack was opaque — I didn't hear mention if it was a closed source proprietary thing or if it used a standard framework to drive the automation.
Last, the report/analysis looks like it has a lot of useful information, but it isn't a great report in practice. This is the same complaint I have with the Test Analytics product they currently offer. It's kind of hard to explain, but using it every day is just very limiting, sluggish, and hard to pull useful information out of.
The Question Nobody at the Summit Asked: Data Portability
My biggest concern nobody asked about was test portability. The plain text test cases could be exported, but if you need to leave BrowserStack or they exit the market, what happens to your test automation? You lose the investment? Back to manual testing?
This has all the same concerns I have with going with proprietary, vendor-locked, solutions (lack of extensibility/customization, cost) but worse. It feels like you are renting the tests in the BrowserStack AI solution paradigm.
My Honest Take
It was good to see the AI trending data. I found that to be useful, but the summit was front-loaded with a lot of marketing in the first 15-20 minutes about BrowserStack in general before that information was shared followed by essentially a commercial for their AI testing product.
AI testing isn't coming. It's here. I don't see the value in the BrowserStack AI curated suite at this stage. I prefer the flexibility and extensibility of using Claude, for example, to develop our own agents and combining them with other open-source frameworks to create test cases we own. From there we can execute locally or on BrowserStack's TurboScale or Automate infrastructure, with results feeding into their Test Analytics product for historical trend tracking and failure triage.
For other less technical companies with simpler web-properties that want a walled garden ecosystem it may be worth evaluating the BrowserStack suite if they are ok with the limitations.
No doubt, we will see a necessary upskill in our profession similar to when manual testing was replaced by automated testing. Developers are using AI to generate code at a pace that needs to be matched by testing. Quality assurance will return to core fundamentals to be able to leverage AI to write good test cases to ensure solid coverage allowing QA to spend more time on creative bugs found by really understanding the business domain, systems, and trends.