Cubeia, an iGaming platform supplier, has shifted much of its software development to AI-assisted coding during 2026, using Anthropic’s Claude model to increase the pace of product releases for its operator clients. The operational change, which began in April and was substantially implemented by mid-July, matters for gambling technology suppliers because faster deployment cycles can affect quality assurance, client integrations and compliance controls in regulated markets.
The company’s chief operating officer, Stefan Grenstad, said Cubeia’s front-end development had become fully AI-driven by the middle of July, while between 90% and 95% of back-end work was being produced through the same approach. The supplier is not describing the move as a licensing, regulatory or enforcement action, but the changes have relevance for operators that rely on platform providers to maintain stable products, deliver required updates and support market-specific compliance requirements.
Cubeia’s experience also illustrates that adopting generative AI in software engineering is not solely a technology decision. The transition has altered the work expected from developers, increased the importance of code review and created pressure to redesign how quality assurance teams are involved in releases.
Before the change, Cubeia generally completed between four and six releases each month, covering an average of about 50 tickets. Grenstad said the volume of work increased as the AI-assisted model gained traction. In May, Cubeia completed five releases involving 86 tickets, while June saw 10 releases involving 96 tickets.
Those figures suggest that AI tools have allowed the company to process a greater number of development tasks, although the increase also creates a broader governance issue. In gambling technology, frequent releases can offer benefits to operators seeking quicker fixes, product adjustments or market adaptations. However, they can also add complexity for clients whose internal teams must test integrations, monitor changes and ensure that systems remain aligned with regulatory obligations in each jurisdiction.
For suppliers serving licensed gambling businesses, development speed does not remove the need for testing, documentation and clear accountability. Platform outages, faulty wallet functionality, errors in reporting or failures in responsible gambling controls can have consequences for an operator’s regulatory standing. As a result, the commercial value of quicker delivery will depend on whether suppliers can maintain the same standard of reliability as release cycles accelerate.
Grenstad said one of the more significant internal issues was not the technical capability of the AI model itself, but the ability of employees to manage multiple tasks at the same time. Developers can initiate several AI agents working on different assignments, with tasks progressing at different speeds. That changes the role of the engineer from writing each line of code to supervising requests, assessing outputs, prioritising work and arranging review.
According to Grenstad, some developers who find it difficult to switch between contexts have struggled with the revised process. The company is considering limits on the number of concurrent tasks that individuals should be expected to manage in a day.
The issue reflects a wider concern around AI-assisted development. A model may generate code quickly, but human staff remain responsible for defining the assignment, identifying errors, understanding dependencies and approving a change before it reaches production. Running too many tasks in parallel could produce bottlenecks in review and testing, particularly if senior developers become responsible for validating a growing volume of generated output.
Cubeia has identified a balance between the number of tasks that can be started and the number that can be properly supervised. Grenstad said setting up an agent can take only a few minutes, creating the potential to launch a large number of assignments. Yet the organisation must still ensure that another developer, or an appropriate reviewer, is available to assess the resulting code.
That requirement may become more important where a supplier’s software supports functions that are central to gambling operations, including player account management, game integrations, payments, bonuses, reporting and responsible gambling tools. Even where code is initially created by an AI model, operators and suppliers may need to show that relevant controls have been tested and that decision-making remains traceable.
The company has encountered differing reactions among its development staff. Grenstad said there had been resistance from some technical leads, although he added that certain early sceptics had adapted rapidly. One senior developer has used Claude as a “sounding board” when working through technical questions, according to Grenstad, treating the model as part of the problem-solving process rather than simply as a code-generation tool.
That approach may help teams use generative AI for drafting, analysis and iteration, but it does not settle concerns over output quality. AI models can generate inaccurate, insecure or unsuitable code, particularly where requests are poorly defined or where the model lacks full context of an existing system. They can also reproduce patterns that require closer human scrutiny before use in production environments.
Grenstad said Cubeia’s staff broadly support the direction of travel, while maintaining concerns about how quality can remain consistent at a higher operational pace. The company’s experience suggests that management acceptance of AI does not necessarily mean every team member is immediately prepared for the organisational demands it creates.
Training and restructuring have therefore become part of the programme. Cubeia is focusing on prioritisation skills and on helping employees decide which tasks require immediate attention and which can be deferred. The company’s technology leads are supporting team members who are finding the change more difficult, Grenstad said.
The shift is also prompting a reassessment of the relationship between developers and quality assurance specialists. Under a more conventional workflow, code may be handed over from development to a separate QA stage. Grenstad said this handover has become less effective as the volume and speed of releases rise.
Cubeia intends to bring quality assurance more directly into the development process, including planning work before coding begins. Under the proposed structure, developers would remain responsible for a task through to production, with QA forming part of that responsibility rather than operating as a separate downstream step.
The plan is to develop AI-assisted quality assurance alongside AI-assisted development. In practical terms, that could mean testing considerations are built into the scope of a task earlier and that developers, reviewers and QA staff work against the same release process. Cubeia has not provided details on the tools or controls it will use for AI-based testing, nor has it said when all business functions will be covered by AI-supported workflows.
The supplier ultimately intends to extend AI use beyond software development across its wider operations. Such a move could affect functions including planning, documentation and support, though the company has not outlined a full timetable or identified which teams will be included first.
For gambling operators, the development may be watched closely because suppliers increasingly compete on their ability to deliver integrations and updates quickly. Platform providers are under pressure to support launches in multiple markets, accommodate different regulatory requirements and provide technical changes requested by clients. AI-assisted coding could shorten parts of that process, but it could also widen differences between suppliers with established engineering controls and those that rely on automated output without sufficient oversight.
There are additional commercial considerations. More frequent releases can require operators to allocate greater capacity to acceptance testing and deployment coordination. Where a platform serves several clients, a change intended for one operator or jurisdiction may need to be assessed for effects on others. The faster a supplier can produce code, the more important it becomes to manage version control, release notes, rollback procedures and communication with customers.
Regulators have not been identified as participants in Cubeia’s internal programme. Nevertheless, licensed operators remain accountable for the systems they use, even when core technology is supplied by third parties. In markets with technical standards or certification requirements, operators and suppliers may need to ensure that changes are assessed under applicable rules before they are introduced.
The company’s move comes as businesses across the technology sector examine how generative AI can reduce development time. The operational challenge is likely to be particularly acute for software used in regulated sectors, where speed must be weighed against reliability, security and auditability.
Cubeia’s next steps will centre on defining workload limits for developers, integrating QA earlier in the release cycle and expanding training for staff adapting to AI-assisted processes. The supplier will also need to demonstrate over subsequent releases whether its higher output can be maintained without weakening testing standards, while operator clients assess the practical effect on deployment schedules and platform stability.
Looking for somewhere to play? Browse our current no deposit bonus listings, with verified codes and the full wagering terms for each offer.
4.8/5
