.png)
If you've sat through a vendor demo this year, you've heard the pitch. Bots that play every build overnight. Regression passes that finish before your first coffee. QA bills cut to a fraction of what they were.
With budgets tight and teams smaller than they were a year ago, it's easy to see why that lands. Nobody wants to pay for testing they don't need.
So let's be clear up front: we're not anti-AI. We're anti-shipping-broken-games. The real question isn't whether AI belongs in your QA process. It's which jobs you hand it, and which ones you absolutely don't.
What AI is genuinely good at
Credit where it's due. Some QA jobs are a perfect fit for automation, and the tools are getting better fast.
- Regression and smoke testing. Running the same checks on every new build is exactly the kind of repetitive grind machines never get bored of.
- Crash detection and soak testing. Leave a build running for 48 hours and let the tools flag every crash, memory leak and performance dip.
- Map sweeps. Bots can run every corner of a level looking for collision holes, stuck spots and fall-through-the-world bugs far faster than a person can.
- Log triage. Sorting thousands of crash reports and grouping duplicates is a job nobody misses doing by hand.
- Localisation checks at scale. Spotting text that overflows its box in 14 languages is tedious for humans and trivial for tools.
And this is all a good thing. Every hour a tester isn't spending on the same checklist for the fortieth time is an hour they can spend doing what people do best.
Where it falls short
Here's the catch. The things that decide whether players love your game or refund it are mostly the things automation can't judge.
Feel and fun
A bot can confirm the jump works. It can't tell you the jump feels floaty, the recoil is off, or the boss fight drags in its third phase. Game feel is a judgement call, and it takes people who've played a lot of games to make it.
Player chaos
Real players are wonderfully unpredictable. They stack crates you never meant to be stacked, pause mid-cutscene, unplu controllers at the worst moment, and sequence-break your carefully built level in ten minutes. The exploit that goes viral on launch day almost never comes from a scripted path. It comes from human curiosity, and good testers have plenty of it.
Judgement
Finding a bug is only half the job. Someone has to decide how bad it is, whether it blocks progress, and whether it matters to the player at all. Get severity and priority wrong and your dev team burns a sprint fixing the wrong things.
Certification
Platform requirements aren't a simple checklist. Passing cert means knowing how each platform holder reads its own rules, and where games tend to trip up. Miss one and you're looking at a failed submission and a slipped date.
Context
Does a line of dialogue land? Does a joke survive translation? Is a menu actually usable for a player relying on accessibility settings? These questions need people who understand the game, the audience and the culture.
The hidden costs nobody puts in the pitch deck
Automation isn't free just because it doesn't need a desk. Someone has to build the tests, and someone has to keep them working as your build changes every day. A new UI layout or a reworked level can break a suite overnight.
Then there are flaky tests: checks that fail at random, get ignored, and slowly teach the team to stop trusting the results. And every report still needs a person to read it, weed out the false alarms and decide what happens next.
The biggest cost is harder to spot. It's false confidence. A dashboard full of green ticks tells you the game runs. It doesn't tell you the game is good. Plenty of titles have launched to glowing test reports and brutal Steam reviews.
What actually works: people with better tools
The studios getting this right aren't choosing between humans and AI. They're putting experienced testers in charge of the tools. Automation covers the breadth. People cover the depth, the judgement and the fun.
That's how we approach it at Kudos. We explore AI tooling aggressively, testing new tools hard to find out what genuinely earns its place and what's just a shiny demo. But we don't bolt anything onto your project quietly. Any AI we'd use on your game gets discussed with you first, so you know exactly what's being used, why, and what it means for your build.
And behind the tools sits what's always made the difference: a team of 100+ in-house, gaming-native testers who play your game like players and break it like professionals, backed by the cert experience to get you through submission first time.
Five questions to ask before you swap testers for tools
Thinking about leaning harder on automation? Run through these first.
- What will it catch that we're missing today? If the answer is vague, the savings probably are too.
- Who maintains it when the build changes? Budget for the upkeep, not just the licence.
- Who decides what's a real bug? Tools find issues. People decide which ones matter.
- How does it handle cert? If the answer is "it doesn't", you still need people who know the requirements inside out.
- What's the plan when it misses something on launch day? Because eventually, it will.
Want more questions like these? Our 10 Questions To Ask Your QA Provider guide goes further.
Game over for human testers? Not even close.
AI is a brilliant addition to the QA toolkit, and it's only going to get better. But it's a power-up, not a replacement for the player. Tools can test your game. It takes people who get games to make it great.
If you're weighing up where automation fits in your QA plan, or wondering whether your current setup is covering what matters, let's talk. Book a free 30-minute consultation and we'll tell you exactly how we'd approach your project, tools and all.
Find out more about our games QA services and co-development.