So… It’s here – the thing I’ve been hinting at across my past two articles.

If you’ve been following my 2-part series about “The Autonomous Testing Tipping Point”, in both parts I talked a lot about the crossroads we’re at in our industry, between the existing world of quality and testing strategies that revolve around “test automation + exploratory testing”, and a new era where I believe AI can remove friction within those existing testing and quality strategies, amplify Quality Engineers, and free up time to be able to focus more on the work that matters most. I also talked about how quality is evolving beyond correctness now, with this more holistic view of quality finally breaking beyond the testing community bubbles into the C-Level and Board Member conversations, and that autonomy in testing isn’t a switch you can just flick on, but is much more like a spectrum you can navigate through.

Well, SmartBear just made their move on that spectrum, and it’s a pretty big move!

So, what is BearQ?

Well, the short answer is that BearQ is SmartBear’s new fully autonomous testing product.

The longer answer, from my perspective, is that it represents a genuine step forward in what testing tools can do, sitting at Levels 4 and Level 5 of SmartBear’s “Levels of Autonomy” for software development. For context, this is towards the highest end of the spectrum where tools aren’t just executing instructions you’ve given them, but are simulating reasoning seemingly well enough to cover planning and the discovery of behaviours within a system with a level of independence. So, not just reactively, like existing test automation, but proactively by scouring through the product on its own accord.

It’s specifically designed to address two problems I kept circling around my head when writing the earlier articles:

1. As AI-generated code and “vibe coding” become more prevalent, the volume, velocity and abstraction of what we’re shipping is only going up. Traditional testing models were already seen as a bottleneck before AI arrived. Now they feel even more out of step with how fast company execs expect to move.

1. Test automation was supposed to free testers up to do the good stuff – exploring the unknowns. But automation still needs building, maintaining, updating, and when something breaks it needs investigating. So what gets squeezed? Yep – the exploratory testing…

BearQ is SmartBear’s answer to these things – using autonomous AI agents to support Quality Engineers.

A quick recap of the launch event

SmartBear ran a virtual launch event to unveil BearQ, and if you missed it, here’s the gist of what they were saying.

A core point they made was one I actually found quite compelling, and if I’m honest, quite aligned with. Their new corporate positioning that ties all their products together is something they’re calling application integrity, which I’ll get into in a moment, but a broader narrative was this: the software development lifecycle has been fundamentally disrupted… Not incrementally changed, but disrupted.

Applications are no longer solely coded by people – they’re AI-augmented, AI-copiloted, or in some cases, fully AI-generated. And those same applications are increasingly being consumed by agents, not just people. So the thing being built has changed, how the thing is being built has changed, and the thing using what’s built has changed. The current testing models aren’t designed for any of this!

What struck me most watching the launch was that SmartBear weren’t just announcing a new product. They were making a philosophical argument about where the industry needs to go, and the product is the expression of that philosophy.

The mission of application integrity

I want to spend a bit of time on this topic, because I think it’s the most important concept to understand if you want to understand why products and tools like BearQ matters.

SmartBear defines application integrity as: “continuous, measurable assurance that your software just works as intended, with the governance to operate at AI speed at scale”. I know… “just works as intended” might be a niggly read for those of us that define quality more holistically, and have lived a life in testing that mostly revolves around investigating unknowns, risks, unexpectations… and all the unintended aspects of the products. But at the heart of what they are saying is: in a world where your application might have been partly ideated, designed, and written by prompts, where there is no visible, clean documentation, and where the system is effectively a black box full of complexity… Even proving that something “operates as intended” is genuinely hard to do.

So their message resonates a bit with me. In “The Autonomous Testing Tipping Point (Part 2)”, I talked about quality going beyond correctness – about the three lenses of goodness, value and correctness providing a more holistic view of quality, and about integrity being a critical thread that runs through it all. The fact that SmartBear are anchoring their product philosophy around this idea of intent-based quality reflects a real shift in how the industry – especially testing tool vendors – are starting to think.

Their answer to the “where’s the source of truth?” question is an API catalogue and test repo that together form a ground truth for software quality. MCP tools for seamless developer and agent access sit on top of that. BearQ then uses that foundation to run continuous autonomous testing, without needing someone to babysit it… This is what we had hoped test automation would bring – freed up time to explore! However, test automation is simply not that autonomous. Will BearQ be the tool that finally enables testers to not have to worry about scripted checks at all, so Quality Engineers can fully pivot to explore left and right across the whole SLDC or DevOps cycle? Additionally, will this also be the tool that shows why QEs will also need to slightly pivot their roles in providing information into the tool, for it to test more effectively in its autonomous manner (i.e. providing it more context, variables, insights, heuristics, risks, etc), while assessing the quality of its output with the rigour that our exploratory testing skills enables. (PS – I say “slightly pivots” the QE role here because it’s the same skills we’re using, just applied in a different context).

BearQ – hugely promising, how  will it land?

Here’s where I’ll be direct with you, because that’s kind of the point of these series of articles…

I think BearQ is a promising product. What SmartBear is attempting, conceptually, is exactly the thing I think will be helpful in a world where we’re building software products using GenAI. Continuous autonomous testing, built around intent, operating across the spectrum of autonomy – this is the direction of travel, and although there are a few companies discussing upcoming tools of this nature, SmartBear is the first I’ve seen actually being released to the market, proving that this direction of travel is happening now.

Will it be universally accepted? That’s an interesting question 🙂

There are a few things that could make adoption of such a product tricky:

First, there is still a lot of scepticism in the testing community, and the wider software community about GenAI in general. There’s also still a level of fear that autonomous tools mean replacement of people. I think SmartBear have tried to be clear that this is not the case – they clearly advocate that this is a tool that supports and enables quality engineers, but there is naturally still scepticism.

Second, the value of BearQ is deeply tied to the quality of your system of record. If your API catalogue is a mess, or your test repository is patchy, or if your teams don’t have clear intent or context documented somewhere… the tool has less to anchor to, so there’s potential for that “garbage in, garbage out” as ever. (Hence why I think a QE pivot towards assessing and improving these things to provide all that context into the tools is one thing that will really matter).

Third, organisations are not all starting from the same place on that autonomy spectrum. What works for a team already operating at level 3 autonomy looks very different from what a team at level 1 needs to do to get there.

But here’s the thing, SmartBear aren’t pretending those challenges don’t exist. Their portfolio of tools (which include other familiar testing solutions: TestComplete, Zephyr, Reflect, QMetry), is nicely designed to meet you wherever you are, and the placement of the portfolio feels like a nice attempt to solve that “black box” challenge rather than paper over it.

“Doom and gloom” or “bloom and boom”?

I know some people will look at BearQ and feel that familiar anxiety of another autonomous AI system… Another thing that could threaten quality or our roles… Being honest again – I’m not in that camp. Some people might feel annoyed or disappointed that I’m not in that camp, but that’s how I feel after getting the chance to play with the product and discuss it with SmartBear.

Here’s my genuine take: the testing tools that are emerging right now are not replacing QEs. They’re finally giving Quality Engineers the breathing room to do the work that actually elevates them. The scripted, deterministic, repetitive checking? Let the tools have it. That frees up QEs to investigate, to explore unknowns, to think about value and impact, to be the human intelligence layer that gives autonomous systems the context they need to actually work well.

BearQ doesn’t make the QE role smaller. It makes it more visible, more strategic, and more impactful.

That’s not naive optimism. That’s what I’m seeing starting to happen already with AI-assisted tools across the industry. And it’s what I argued in detail in my last article when I talked about QEs as curators of intent, designers of constraints, interpreters of behaviour, etc. BearQ is a product that needs exactly those skills to function at its best!

Zooming out… What SmartBear is building around application integrity feels like it could genuinely shift the conversation at a level that matters. Yes, I think we need to define “integrity” to encapsulate quality beyond “correctness” too, but the philosophy of it all remains the same.

Right now, a lot of organisations just plain and simply look stuck. Cost cuts have hit QE and testing teams hard, while complexity is growing within software products. Delivery expectations aren’t slowing down – in fact they’re increasing, adding more and more stress and strain on the operational systems. And there’s this growing awareness at exec and board level that “correctness” isn’t sufficient as a quality bar anymore.

What the industry seems to need is a path forward that doesn’t require organisations to blow up their existing toolchains or retrain entire teams from scratch. It seems to need something that integrates and scales, and supports quality (holistically) in being measurable at the speed in which AI tools are being used to create products.

I’m still having a play… and I’ll report back on that soon

I mentioned that I’ve been given early access to BearQ, and I’ve hooked it up to my prototype “impact driven growth framework” web-app that I vibe-coded (so putting it through its paces 😂).

I’m still playing – but what I can say is that the early experience is compelling. In my next article, I’ll give you my honest assessment – what impresses me, what surprised me, what questions I still have and what improvements I’d suggest (no sponsor-friendly softening 😉). Just what I actually think after having actively used it over a longer period of time.

Watch this space…


This article is supported by SmartBear. They provided me with early access to BearQ, so that I could explore it and share my honest views and experiences of the tool and their philosophy around it. If you’re interested in learning more about BearQ and requesting early access, visit this page.