Side, a long‑standing services provider, has entered into a six‑month memorandum of understanding with Denmark‑based artificial‑intelligence specialist Modl.ai to co‑develop an AI‑driven quality‑assurance platform for video games. The agreement binds the two parties to collaborate on a system that blends Side’s extensive QA expertise with Modl.ai’s cutting‑edge machine‑learning capabilities, aiming to create a more efficient, scalable testing solution while keeping human oversight at the core. Side, previously known as PTW, has been a fixture in the outsourcing and testing market for years, offering a broad suite of services to game developers worldwide.
Modl.ai, by contrast, is a newer entrant that launched in 2018 and secured $1.7 million in seed financing two years later. The partnership is presented by Side’s chief commercial officer, Martin McBride, as a union of complementary strengths: “The MOU brings together the best of Side and the best of Modl.ai into a single, differentiated offering. One of the holy‑grail aspirations for the games industry today is to harness AI to augment QA services. No one has cracked that problem yet, and we believe we have a viable answer.” Earlier AI‑enabled QA tools—some of which were also created by Modl.ai—required developers to ship entire builds or even source code to the testing vendor.
Those solutions relied on deep integration via plugins and SDKs, which introduced significant friction and, as Modl.ai founder Christoffer Holmgard admits, became “difficult over time.” The new approach eliminates the need for any internal access. Instead, developers provide the product exactly as they would hand it to a human tester or eventual consumer, install it on a comparable device, and let the AI run the game, watching the screen, simulating inputs, and recording every observable event. A second AI layer then analyses the generated data, flags potential defects, and escalates the findings to a human reviewer for final validation.
The first phase of the workflow is what Modl.ai calls “building game knowledge.” Developers upload assets—textures, models, design documents, even screenshots—and converse with the AI in natural language to teach it the context and mechanics of their title. Holmgard explains, “All of this happens in plain English.
We chat with the system about the game, feed it images, design docs, and other artefacts, so the AI builds a mental model of what the game is supposed to do.” Once the AI has a baseline understanding, developers specify the testing objectives. These can range from traditional test cases (e.g., “verify that the main menu loads within three seconds”) to novel scenarios that are deliberately crafted to exploit the AI’s pattern‑recognition strengths. “Human expertise is still required to phrase the prompts correctly and guide the AI toward the right behaviours,” Holmgard notes, emphasizing the collaborative nature of the process. During a test run, the system captures a wealth of evidence: textual logs, short video clips, and comprehensive telemetry.
“We receive the raw text, video snippets we call ‘evidence,’ and all the associated logs,” Holmgard says. This package is handed to a seasoned QA specialist who reviews each report, decides whether the flagged issue is a genuine bug, a design feature, or a false positive, and provides feedback that the AI can learn from. Over time, this iterative loop refines the model’s ability to distinguish subtle nuances between intentional gameplay elements and unintended defects. One of the perennial challenges in game QA is discerning a bug from a feature, especially in titles with emergent or sandbox mechanics.
Holmgard points out that continuous feedback from human testers helps the AI calibrate its judgments: “Prompt engineering, pre‑built game knowledge, and ongoing human feedback all combine to make the system more attuned to the specific design philosophy of each game.” Side envisions a broader impact on the labour market. By upskilling QA staff into AI engineers, the company hopes to elevate the value chain and potentially repatriate jobs that have migrated to low‑cost regions. “Our people can move from being pure testers to AI engineers, earning higher salaries and opening new career pathways,” McBride explains. This shift also promises a more fulfilling work experience, as repetitive, low‑skill tasks are automated, freeing testers to focus on strategic analysis, creative exploration, and higher‑order problem solving.
Modl.ai’s chief product officer, Benedikte Mikkelsen, echoes this sentiment: “Many current tests are tedious and can be run by a bot. That frees humans to evaluate gameplay, devise inventive test scenarios, and add real value.” Both executives stress that AI is not a replacement for human QA but a tool that amplifies human capabilities and reduces error rates. Regarding the types of games the solution currently supports, Side’s chief operating officer, Oliver Deslandes, is candid: the system works best with simpler titles—puzzle games, card games, and other experiences with repeatable mechanics—rather than fast‑paced 3D open‑world or fighting games. “We’re targeting projects that have clear, repeatable loops.
More complex, high‑velocity games will come later as the technology matures.” Deslandes adds that early adopters will likely be developers willing to experiment and pilot the technology, accepting that the journey will involve iterative refinement. In summary, the Side‑Modl.ai partnership aims to create a hybrid QA model where AI handles bulk, repetitive testing and data collection, while seasoned human specialists provide the critical judgment and domain expertise that machines lack. By treating AI as a “sliding scale” with a permanent human in the loop, Side hopes to raise testing standards, reduce costs, and open new professional avenues for its workforce, all while navigating the current technical limitations of AI in the fast‑evolving gaming industry.