In an era of rapid digitisation and rising customer expectations, the traditional view of quality assurance as a technical checkpoint is giving way to something far more strategic.
For Applitools, a major player in visual testing and AI-assisted automation, that evolution has been years in the making, and it’s now accelerating with the emergence of purpose-built AI models.
“AI is a catch-all term,” said Kate Cato, since May 2024 the Chief Customer Officer at Applitools, in a recent conversation with QA Financial. “But the models that matter in QA are purpose-built, secure, and efficient. And that makes all the difference in environments like financial services.”
Asked about the role of deterministic AI in future testing strategies, Cato pointed to a philosophical divide in the market.
“There’s a lot of noise out there. Most platforms are wrapping general-purpose LLMs with thin interfaces. That might look impressive, but it’s not built for the realities of enterprise testing, especially not in regulated industries like banking,” she said.
Applitools has taken a different path: developing in-house, domain-specific models that are deterministic and tightly scoped.
“We don’t need our testing tool to know if Hungary had a president 50 years ago. We need it to write the right command steps, validate workflows, and ensure accuracy at every layer of the stack.”
Security, predictability, and data protection are non-negotiable for Cato and her company. “Our clients know their information won’t leak into the public sphere. That’s a critical differentiator for banks,” she stressed. “Everything we build, from our computer vision engine to our deterministic LLM, is developed with that in mind.”
All eyes on Autonomous
QA Financial asked how Applitools’ approach helps QA teams move beyond fragile, selector-based testing, especially important in banking environments with fast-changing UIs and legacy systems.
“We never wanted to just make existing tools a little easier,” Cato explained. “From the beginning, with our Eyes platform, we didn’t say, ‘Let’s make assertions simpler.’ We asked, ‘Why are we even using assertions for visual problems?’ We built a visual model to solve a visual issue.”
That same philosophy now underpins Applitools Autonomous, the company’s relatively young LLM-powered test generation tool. Rather than rely on brittle DOM selectors that can break when attributes change, the Autonomous system interacts with the page using language, not just code.
“The power of an LLM is that it actually reads the page and understands the intent,” Cato noted. “So when a test fails, it’s not because a CSS selector changed, it’s because something meaningful happened. Maybe the login button disappeared. That’s a failure you want to catch.”
And it’s working. “We track the failure rates,” she said. “With Autonomous, flaky tests due to element changes are almost nonexistent. When a test fails, it’s because something important actually changed in the application.”
Testing across environments
One of the most significant time-savers for financial services firms, Cato continued, is the ability to design tests that can run consistently across pre-production, staging, and production environments.
“Being able to write a test once and use it across environments, that’s where efficiency shows up. It reduces duplication, lowers maintenance, and ensures consistency. It also helps teams catch issues earlier with more confidence.”
The use of parameterised data also plays a key role. “In banking, you’re often working with mock customer or account data for compliance. Having pre-approved test variables means testers can move quickly without needing direct access to sensitive systems,” she said.
Perhaps the most powerful shift has been how AI is reshaping who participates in testing. Autonomous testing, especially with natural language prompts, has made it possible for marketers, designers, and product managers to write and run meaningful tests without deep engineering support.
“With the latest release, users can just type something like, ‘Log into the banking app,’ and our model will break it down into human-readable steps: enter email, enter password, press submit,” Cato pointed out. “It’s clear, it’s readable, and anyone can understand it.”
“We don’t need our testing tool to know if Hungary had a president 50 years ago.”
– Kate Cato
Test steps are aligned with visible elements, not hidden technical selectors, which boosts both accuracy and accessibility.
“If something’s off in the generated test, it’s easy to fix,” she added. “You can replace a suggested name with a secure test user variable from your organization’s approved data set. That’s where our work on variables and secure inputs really pays off.”
Asked whether testing is finally being seen as a strategic enabler rather than just a cost center, Cato was unequivocal.
“Yes. And the disruption here is coming faster than in most sectors. For years, QA was essential but underfunded. It was seen as important, but not innovative. That’s changing.”
She pointed to Applitools’ long-standing investment in proprietary computer vision, technology that’s delivered measurable efficiency gains for enterprise teams. “We’ve consistently seen 30% improvements in test coverage and stability. With LLMs and AI agents now in play, we see another 30% gain on the horizon.”
But it won’t come from one-size-fits-all models. “General-purpose AI is too vague, too expensive, and too insecure for what banks need. We’re proving that targeted AI works, and we’re building toward that next layer of embedded, real-time testing.”
Looking ahead
Before wrapping, QA Financial asked Cato what lies ahead for Applitools, and for quality assurance more broadly.
“We’re really excited about what’s coming,” she said. “The expansion of Autonomous to mobile is a big one. Mobile testing has always been difficult, but with computer vision and LLMs working together, we’re unlocking it for a much broader range of users.”
Applitools is also investing heavily in component-level testing, especially for a world shaped by tools like GitHub Copilot.
“As developers begin to use AI to write components, we’re asking: how do we build testing into that flow, in real time, as the code is being written? So that when it hits the client, it’s already been validated.”
Cato concluded by saying that “there’s a lot of uncertainty in the AI space right now, especially around regulation and security. But our mission is clear: to build trustworthy, efficient testing systems that work at enterprise scale, especially for the industries that can’t afford to get it wrong.”
NEW EVENT
Why not become a QA Financial subscriber?
It’s entirely FREE
* Receive our weekly newsletter every Wednesday * Get priority invitations to our Forum events *
REGULATION & COMPLIANCE
Looking for more news on regulations and compliance requirements driving developments in software quality engineering at financial firms? Visit our dedicated Regulation & Compliance page here.
READ MORE
- Leapwork engineering head: Why test automation so often fails to deliver
- World Economic Forum warns financial sector must strengthen AI risk controls
- How Banca Progetto is hard-wiring quality into Italy’s digital banking space
- How Wealthsimple builds quality into the product, not around it
- Digital revamp puts spotlight on internal controls at China Construction Bank
WATCH NOW
