Cubeia, a Swedish software company, is in the third phase of its journey towards AI-assisted development, evolving from letting developers use AI as they wish to establishing a structured AI-driven development process. Now, it confronts a vital question: how should the organization operate when coding has become less of a bottleneck?
"We’re doing this 100%,” said COO Stefan Grenstad regarding the implementation of the AI-driven development process. The focus has shifted to how Cubeia can leverage the additional capacity that AI has enabled. Just six months prior, the company's experiments mainly involved substituting human-written code with AI-generated alternatives.
The current approach encompasses a redefined development pipeline, adjusting the roles of developers and quality assurance (QA), reorganizing teams, and fostering a fresh relationship with clients.
Cubeia's initial phase featured an open AI approach, which allowed developers to utilize the technology freely. The subsequent phase introduced structure, mandating that all developers operate under the same agents and AI-driven pipeline. This restructuring required Cubeia to tackle issues concerning quality, reliability, and collaboration among agents, while helping employees transition to a new working model. Grenstad asserts that this adjustment has been largely accomplished.
“During the hybrid period in Q1, Cubeia solved 259 issues. After migrating to the AI-driven process, that number jumped to 421—a 62% increase. Larger projects surged from 17 to 58,” Grenstad reported, emphasizing that transitioning to AI-infused development has greatly enhanced output.
A key focus now is what the company can achieve with this newfound capacity. "Because we’re not dedicating as much time to coding, we’re now concentrating more on the business: discussing value and comprehending the domain,” Grenstad noted.
Instead of merely fulfilling requests, teams are increasingly expected to grasp the rationale behind challenges and the value they ought to create. Paul Crisp, Cubeia’s marketing head, elaborated, "If the goal is to achieve 10% more traffic, we need to establish a baseline, implement the change, measure it, and verify whether it indeed increases by 10%. If it only rises by 1%, we may need to iterate further to accomplish the goal.”
Customer input is driving the next stage of development. Cubeia is now engaging clients by asking, "What does this mean for the customer? What can the customer actually achieve with this?" Recently, a client requested to let their AI agent manage data streams using Cubeia’s platform.
One client created its own casino landing page, using Cubeia's APIs for integration. "Another wanted to use our player account management system to build unique functionalities like personalized bonuses,” Grenstad recounted, highlighting the shift towards allowing customers to access Cubeia’s data and platform to develop their own AI-based solutions. A pilot project is currently underway based on the landing-page example.
Just half a year ago, such an initiative seemed unlikely, but Grenstad is now confident that Cubeia should be prepared to “say yes” to clients presenting products built with their own AI tools. "We’ve realized we’re going to be part of it,” he remarked.
The heightened capacity is transforming Cubeia's team structure as well. The company is trialing a rapid-response team focused on handling smaller customer requests alongside another team dedicated to larger projects and long-term goals. Current teams consist of five to eight members, but Grenstad believes they could ultimately be streamlined to two or three-person units. Developers are also free to transition between rapid-response and longer-term projects based on their interests and optimal contributions.
However, increased capacity does not eliminate all bottlenecks. With AI agents handling multiple streams of work simultaneously, the challenge lies in determining how much work an individual can oversee and review. "The bottleneck becomes the person reviewing everything. We still examine code in sensitive areas, especially concerning architecture, performance, security, and reliability, but we no longer conduct extensive manual reviews for all code,” Grenstad noted.
As AI influences recruitment strategies, teams focused on the player experience need individuals who grasp aspects of casinos and iGaming. Grenstad suggested, "I would not necessarily prioritize senior Java or front-end specialists the same way as before. While technical skills are still important, domain knowledge has become even more critical. I’d prefer a candidate with strong domain experience over one who simply knows Java but lacks industry insight."
This evolution raises questions about future industry entrants. With the demand for traditional junior Java developers diminishing, the conventional trajectory from junior to senior developer becomes less defined. Grenstad expressed concern about ensuring that younger talent enters the workforce over time, raising important discussions within the company.
Cubeia is also transitioning from the traditional development process where a developer creates something and passes it to QA. In the past, Grenstad explained, the product owner determined what needed to be built, developers figured out how to build it, QA performed testing, and operations managed the release. Now, QA becomes an integral part of the process, with product engineers contemplating automation and manual testing from the planning stage. QA professionals support automation while identifying coverage gaps, extending accountability from problem identification to implementation and measuring outcomes.
Despite this transformation, challenges remain. Some coders find themselves missing the hands-on coding tasks they once enjoyed, as those elements of their roles have diminished. Additionally, there are strategic concerns about reliance on AI providers, especially with Cubeia's strong dependence on Claude. Grenstad acknowledged that potential price hikes or changes in quality from AI providers remain uncertainties, but he also views the experience as an opportunity for Cubeia to learn and adapt.
"I have confidence that we can manage the unknowns as we continue our journey. We’ll confront emerging challenges, learn to handle them, and adjust accordingly,” Grenstad affirmed.
Reflecting on Cubeia's early ambitions, Grenstad remarked, "When we began this journey, my only vision was: ‘Let’s not be writing code in August.’ We set out to eliminate coding as the barrier to software development but discovered that coding was just one limitation in a far more complex landscape. As the bottleneck shifts, so too do the demands placed on the organization, including expertise in domain knowledge, product focus, quality, prioritization, and understanding customer needs. Ultimately, the technology has taken a back seat in this narrative, while the emphasis now rests on how Cubeia will utilize that technology moving forward.
