Side, a seasoned services provider, has entered into a six‑month memorandum of understanding (MOU) with Denmark‑based artificial‑intelligence specialist Modl.ai. The agreement binds the two organisations to co‑develop a quality‑assurance (QA) platform that leverages AI while retaining human oversight.
Side, previously known as PTW, has long been entrenched in the services sector, offering a range of support functions for game developers. Modl.ai, by contrast, is a relatively new player in the gaming arena. Founded in 2018, the startup secured $1.7 million in seed capital two years later and has since focused on applying machine‑learning techniques to the testing of interactive entertainment. "The MOU essentially merges Side’s deep QA expertise with Modl.ai’s cutting‑edge AI capabilities into a single, differentiated offering," explains Martin McBride, Side’s chief commercial officer.
"One of the holy‑grail ambitions for the games industry today is to harness AI to augment QA services in a meaningful way. No one has cracked that problem yet – we believe we are on the cusp of doing so." Earlier AI‑driven QA solutions, including earlier versions from Modl.ai, required developers to ship full builds or even share source code with the testing firm.
Those approaches relied on integration plugins and software development kits (SDKs) that created friction and, over time, proved cumbersome, according to Modl.ai founder Christoffer Holmgard. The new workflow eliminates the need for any internal access. "We treat the product exactly as you would hand it to a QA tester or eventually to a consumer," Holmgard says. "You install it on a device that is either identical to or very similar to the target hardware.
Our system then runs the game, watches the screen, simulates key inputs, records every observable event, and passes that data to a second AI layer that flags potential bugs. That layer then escalates the findings to a human reviewer for validation." The first phase of the process is what Modl.ai calls "building game knowledge." Developers upload assets—textures, models, design documents, even screenshots—and converse with the AI in plain English (or other natural languages).
The AI assimilates this information, creating a contextual model of the game’s mechanics, narrative flow, and visual style. "All interactions are in natural language," Holmgard notes.
"We might chat about the game, attach images, or feed in design specifications. The AI learns from those inputs and builds an internal representation that it can later use to generate test scenarios." Once the knowledge base is in place, developers specify the testing objectives. These can range from traditional test cases—such as verifying menu navigation or collision detection—to more exploratory scenarios that are uniquely suited to AI‑generated probing. Here, human expertise is still essential.
"You need skilled people to craft prompts that steer the AI in the right direction," Holmgard explains. "The quality of the output depends heavily on how well you phrase the instructions and on the depth of the game knowledge you have already supplied." During a test run, the system records logs, captures short video clips as evidence, and compiles textual bug reports. These artefacts are handed to a human QA specialist, who reviews them, decides which issues merit action, and feeds the decisions back into the AI to refine its future judgments. "Distinguishing a genuine bug from an intended feature can be subtle," Holmgard admits.
"Human feedback helps the AI learn those nuances. Over time, as testers label reports as ‘bug,’ ‘not a bug,’ or ‘design choice,’ the model becomes increasingly attuned to the specific design language of that game." Side envisions a longer‑term shift where QA personnel evolve into AI engineers.
By upskilling testers to manage and fine‑tune the AI tools, the company hopes to raise the value of its workforce, repatriate jobs to regions that have lost testing roles due to cost pressures, and create new career ladders. "One of the exciting outcomes of programs like this is that we can move our people up the value chain," McBride says. "Instead of a linear path from junior tester to senior tester, they can become AI engineers, earn higher salaries, and open up fresh career trajectories.
This also gives us the chance to bring jobs back to markets where they have historically disappeared because of low‑cost labour dynamics." McBride adds that AI should make QA work more rewarding, not replace it. By automating repetitive smoke tests and routine verifications, testers can focus on strategic analysis, creative exploration, and higher‑order problem solving. "The tool excels at eliminating the drudgery of repetitive checks," he notes. "When each new build arrives, there are countless smoke tests and verification steps that can bog down a tester’s day.
Automating those frees the team to think strategically, conduct deeper analysis, and ultimately enjoy a more fulfilling career at Side." Benedikte Mikkelsen, Modl.ai’s chief product officer, echoes the sentiment: "Many of the current tests are tedious and involve running the same flow over and over. An AI can take over that monotony, allowing humans to concentrate on assessing gameplay, crafting innovative test scenarios, and adding real value. That’s the core proposition of AI‑enhanced QA." Both executives stress that AI will never fully replace human QA staff. The price point for low‑cost testing in certain regions remains too low to eliminate the need for human involvement entirely.
Instead, AI acts as a force multiplier, extending the capabilities of existing teams. "There will always be a place for people to do testing work because the economics of ultra‑low‑cost testing are still a reality," McBride says.
"But by providing sophisticated toolsets, we empower our staff to advance their careers, stay longer with the company, and move up the value chain." Side’s philosophy frames AI as a "sliding scale"—a spectrum where human oversight is constant, but the degree of automation can be adjusted based on project needs. "Our ethos is that there will always be a human in the loop," McBride asserts. "The AI gives us the best of what the technology can offer—speed, consistency, and error reduction—while humans provide judgment, creativity, and contextual understanding." When asked about the types of games the solution currently supports, COO Oliver Deslandes is candid: the system is best suited for simpler titles—puzzle games, card games, and other experiences with repeatable mechanics—rather than fast‑paced 3D open‑world or fighting games. Those more complex genres will require further development.
"We’re focusing on games that aren’t too rapid, where mechanics are repeatable and the visual feedback is manageable for the AI to interpret," Deslandes explains. "Clients willing to embark on this journey with us will see incremental improvements as we refine the technology. Some will demand immediate cost savings; others will be more interested in long‑term innovation." He concludes, "It’s a journey. Early adopters who run pilots with us will get a front‑row seat to watch the technology evolve and understand where it can take their QA processes."