.svg)
When you think about test automation, what image comes to mind?
For many QA leaders, automation still means running the same scripts every night, chasing down false-positives, and fighting maintenance debt. That model served us well for a while, but it was always limited: automation only runs what humans script.
The next era is about AI-powered testing. AI-powered testing doesn’t just execute predefined tasks; it generates coverage dynamically, adapting as your application evolves.
When I sat down with @Alex Martins, our VP of Strategy, the first thing we tackled was the hype around AI. Everywhere you look, headlines promise transformation, but the reality is that most enterprises haven’t seen a single percentage point of real business value. Recognizing this gap between hype and results is the first step toward making smarter decisions about your testing strategy.
Here's the full video of my conversation with him:
The difference between traditional and AI‑driven testing
Traditional automation has always been human‑driven.
Someone writes a script, records a flow, or drags and drops components in a low‑code tool. The machine then executes exactly what it’s told. Think about it: humans manually creating automation. That’s not innovation, it is busywork. AI‑driven testing flips that equation. Instead of relying on brittle scripts, you let AI observe real user behavior and generate tests automatically.
This change redefines QA. When AI monitors your end users activities in production, it sees which paths users actually follow through your app, not just the paths defined in a requirements document. As Alex explained, AI can take inputs from end users in production, notice their behaviors and flows, and then automatically generate tests that truly represent what matters to your end users and, therefore, to your business. It also adapts when a change happens in the application, suggesting new tests to cover those fresh journeys.
In pre‑production, AI observes all testers validating new features across positive scenarios, negative scenarios, edge cases, etc. based on requirements and then converts those manual runs into automated tests that can be immediately integrated into your CI/CD pipeline.
You’re no longer stuck writing and maintaining scripts; the machine does the heavy lifting, while you focus on whether the tests make sense for your users and your business.
Building on real user behavior instead of assumptions
One of the most compelling benefits of AI-powered testing is its foundation in real user behavior.
Back in my product management days, I remember how often we’d fall into the trap of designing and testing only the ‘happy path.’ Step one, click here. Step two, fill in this field. Step three, everything works as planned.
But reality rarely matched the script. Users would take unexpected turns, and entire flows we never anticipated would surface. That gap between what we imagined and what users actually did is exactly why AI-native testing is so powerful, it closes that blind spot.
Image: How TrueTest analyzes production environment to find coverage gaps
TrueTest captures what people actually do. It maps user journeys across your application: which pages they visit, how they navigate, where they encounter friction. This doesn’t just reduce wasted effort on edge cases no one touches; it reveals hidden gaps and alternative paths you never anticipated.
When QA teams see that their automated tests don’t match user behavior, they can prioritize coverage where it matters most and raise strategic concerns to product owners.
The impact extends far beyond quality assurance. Product managers can use these insights to understand which features users ignore, or which ones they use the most. Business analysts can identify unmitigated risks when users wander into unintended flows. QA moves from being a back‑office function to a strategic partner, providing data that shapes product roadmaps and user experience decisions.
AI is not magic
Despite the promise of AI‑driven testing, cultural and process shifts are essential. One question I asked Alex was how teams move from “running scripts” to letting AI generate them. Many testers still cling to manual processes because they feel in control.
Simply handing them an AI tool and expecting instant adoption doesn’t work.
Alex noted that a lack of training is a major barrier to productive AI use. Executives may embrace AI, but teams on the ground often resist new technology that potentially changes the way they work. It's normal human behavior.
The fix comes down to two things: integration and trust. Integration means embedding AI into the workflows testers already use, not forcing them to relearn everything from scratch. At Katalon, we infuse AI into every step of the software testing lifecycle, guiding testers through the process so results are consistent no matter who’s behind the keyboard.
Trust comes from education and enablement. Testers need to see how AI helps them incrementally at first, then they will feel how the compounding effect of those incremental improvements gives them super powers. When they understand it, AI stops being a black box and starts being a partner.
Testing for AI agents
There’s a new challenge rising fast: AI agents are now using applications on behalf of humans. We’ve already seen it firsthand when some of the registrations for a recent Katalon webinar were executed by ChatGPT, not people.
And it’s not just sign-ups. Agents are booking flights, making online purchases, and navigating experiences in ways humans don’t. They take different routes, click different elements, and interpret instructions literally. If your tests only cover human flows, you’re blind to agent flows. And every failed interaction means lost revenue.
TrueTest closes that gap by capturing both human and agent behavior, then generating tests that mirror each journey. That’s critical as agent adoption accelerates. Early evidence shows agent traffic is already becoming a meaningful share of usage.
The takeaway is clear: if you want to protect customer satisfaction and revenue, you need to test for both. This isn’t a future problem. It’s here.
Testing what really matters
Another misconception about AI‑generated tests is that they guarantee “full coverage.” When I asked Alex whether the approach would provide full coverage for any app, Alex clarified that it depends on our definition of “coverage”.
TrueTest aims to provide appropriate coverage from a requirements and real‑user perspective. The goal is not to exercise every line of code; it’s to ensure that the experiences users have in production are free of issues.
You can achieve 100‑percent code coverage and still release a buggy product. No single metric should stand alone. AI-powered testing combines requirements coverage, user journey coverage, and (where appropriate) code coverage to minimize risk.
Practical next steps for QA leaders
The discussion closed with a practical takeaway: explore AI capabilities that incrementally improve existing workflows. If you’re a QA leader or product owner, you don’t need to overhaul your entire testing process tomorrow. Instead, start by:
- Identifying high‑traffic flows in your application. Let AI observe user behavior and generate tests for those flows.
- Comparing AI‑generated tests to your existing scripts. Notice where your assumptions about user behavior differ from reality and let AI close the gap for you.
- Investing in training. Make sure testers understand how to use AI tools effectively and validate their benefits.
- Including product and business teams in the conversation. Share user journey insights and discuss how they impact feature prioritization.
- Experimenting with agentic scenarios. As AI agents become more common, begin capturing and testing for those interactions to ensure AI agents can achieve their goals on your websites.
AI‑driven testing is not about hype; it’s about practical intelligence. It’s about augmenting human testers, aligning quality efforts with user value, and turning QA into a strategic advantage. By grounding test generation in actual behavior, reducing maintenance overhead, and embracing continuous feedback, you can deliver software at AI speed without sacrificing confidence.
The future of quality isn’t just more automation. It’s intelligence. It’s a shift in mindset from reactive script running to proactive risk detection. And it’s happening now.
