Side, a seasoned services provider formerly known as PTW, has entered into a six‑month memorandum of understanding with Denmark‑based artificial‑intelligence specialist Modl.ai. The agreement sets the stage for a joint effort to build a next‑generation quality‑assurance (QA) platform that blends AI automation with human expertise.
While Side brings decades of experience in testing services and a global workforce, Modl.ai contributes cutting‑edge machine‑learning technology and a fresh perspective on how games can be examined for bugs and design flaws. The partnership is rooted in a shared belief that AI should function as a "sliding scale" rather than a wholesale replacement for people.
Martin McBride, Side’s chief commercial officer, describes the collaboration as a marriage of the best assets each company possesses: Side’s deep‑rooted QA processes and Modl.ai’s innovative AI algorithms. He calls the venture "the crown‑jewel, holy‑grail experience" for the games industry—a solution that finally lets developers harness AI to augment, not replace, traditional testing. Historically, AI‑enabled QA tools required developers to hand over entire game builds, source code, or integrate cumbersome SDKs. Modl.ai’s earlier offering suffered from exactly that friction: developers had to grant the AI system deep access to internal assets, which proved impractical for many studios.
The new approach, however, eliminates the need for any internal code exposure. Instead, the AI treats the game exactly as a human tester would—by running the compiled product on a device that mirrors the end‑user environment. The system records visual output, simulates inputs, and compiles a comprehensive data set of every in‑game event. A second AI layer then analyses this data, flags potential defects, and escalates them to a human specialist for final judgment.
The first phase of the workflow is what Modl.ai calls "building game knowledge." Developers upload assets, design documents, screenshots, and other relevant materials, then converse with the AI in plain English (or any natural language). This dialogue helps the system construct a semantic model of the game's mechanics, narrative, and visual style. By feeding the AI both textual descriptions and visual examples, the model can later interpret test results in a context that aligns with the developers' intentions.
Once the AI has a working understanding of the title, the development team defines the testing scope. This can range from conventional test cases—such as verifying menu navigation or ensuring that a level loads within a target time—to more exploratory scenarios that are specifically crafted to exploit the AI's pattern‑recognition capabilities.
Holmgard emphasizes that human insight remains crucial at this stage; testers must phrase prompts in a way the AI can act upon, effectively teaching the system how to think like a QA professional. During a "run"—Modl.ai’s term for a testing session—the AI executes the prescribed actions, captures video snippets, logs system events, and generates textual reports.
These artifacts, collectively referred to as "evidence," are handed over to a human QA analyst. The analyst reviews the findings, decides whether each flagged issue is a genuine bug, a feature, or a false positive, and provides feedback that the AI incorporates into its learning loop. Over time, this iterative process refines the AI’s ability to distinguish subtle design nuances, reducing the rate of misclassifications.
One of the most challenging aspects of automated QA is discerning intent: is a character’s unexpected movement a glitch or a deliberate gameplay mechanic? Continuous feedback from human testers helps the AI calibrate its judgments. As Holmgard explains, the system draws on three sources of knowledge: the initial prompting, the pre‑built game model, and ongoing human annotations. This triangulation enables the AI to become increasingly attuned to the specific design language of each project.
Side sees the technology as a catalyst for career evolution within its QA workforce. By automating repetitive, low‑value tasks—such as smoke tests for every new build—testers can focus on higher‑order analysis, strategic problem‑solving, and creative exploration of game mechanics.
McBride envisions a future where QA professionals transition into "AI engineers," overseeing the prompts, interpreting AI outputs, and continuously improving the testing pipeline. This shift not only raises individual earning potential but also opens pathways to repatriate jobs to regions where they were previously outsourced due to cost pressures. Benedikte Mikkelsen, Modl.ai’s chief product officer, echoes this sentiment, noting that many current tests are monotonous and could be performed more efficiently by a machine.
By freeing human testers from rote actions, they can devote their expertise to evaluating gameplay balance, narrative cohesion, and other nuanced aspects that remain beyond the reach of current AI. Despite the enthusiasm, both companies acknowledge the technology’s present limits.
Oliver Deslandes, Side’s chief operating officer, admits that the system currently excels with simpler titles—puzzle games, card games, and other experiences with repeatable mechanics—rather than fast‑paced 3D open‑world adventures or high‑intensity fighting games. The focus, for now, is on clients willing to pilot the solution and help shape its evolution.
Early adopters can expect a collaborative journey, with incremental improvements as the AI learns from real‑world usage. In summary, the Side‑Modl.ai partnership aims to create a balanced QA ecosystem where AI handles the bulk of repetitive testing while humans retain ultimate authority over defect validation and strategic decision‑making. By treating AI as a scalable tool rather than a replacement, the alliance hopes to deliver higher‑quality games, more satisfying careers for QA staff, and a flexible testing framework that can adapt to the diverse needs of the gaming industry.