High stakes: AI and Machine Learning risks and opportunities in igaming
AI is here to stay
AI and ML are far more than buzzwords, these technologies have already been part of the reality of online gambling for many years. Igaming platforms mainly use AI to provide customers with targeted advertising, personalised promos and bonuses, customised game and bet offers, tailored customer support and to help with responsible gaming requirements.
Harnessing AI and ML to create personalised betting and gaming experiences is a logical progression not only for operators but also for players. Research found that 80% of igaming customers rated personalised offers and bets as “valuable” or “very valuable”, and 75% found non-personalised apps difficult to navigate.
Starostenkov is already implementing his vision at EvenBet Gaming to capitalise on these trends but remains realistic about its challenges.
“However, while it is a tough task, EvenBet consistently monitors AI technology and trends to keep our player protection and support tools updated to the highest possible standard, ensuring our operators’ reputations are safeguarded, and revenue is shielded from fraudulent activities.”
Managing risks
Although AI and ML are causing a paradigm shift in igaming by creating new avenues for innovation and driving growth and user engagement, these advancements also pose risks that technology developers, operators and players must be aware of and manage carefully.
Starostenkov finds similarities across how operators and suppliers currently implement ML technology: “Both operators and suppliers are actively working toward a similar goal, attempting to decipher how technology can help us solve the intricate problems of responsible gaming, fraud, and finance control.”
Battling the bots
Bots that artificially imitate players have been a problem the igaming industry has been facing for years. Whether taking unfair advantage of human players on a peer-to-peer online poker platform or exploiting an online casino to generate fraudulent winnings in slots and table games, bots severely undermine the fairness and integrity of online gambling.
“The obvious and most persistent danger comes from bots, which utilise AI and machine learning to imitate human behaviour in a very realistic way,” explains Starostenkov. “The duration of their game sessions varies, they do not showcase repetitive patterns, and they even participate in chats, making it incredibly difficult to identify them. For casinos, winning bots result in considerable financial damage, and for online poker rooms, a potential loss of player trust too.”
AI-powered bots, especially when used in skill-based games like poker, can analyse game patterns, predict odds and outcomes, and optimise and execute decisions more quickly and efficiently than human players. As they are also much harder to detect, EvenBet Gaming tackles the detection with a measured approach.
“We recommend executing a hybrid approach, implementing AI technology for identification and signalling while retaining humans for expertise in applying restrictive measures. AI is a powerful tool for identifying suspicious behaviour, and we have reaped the rewards of these benefits on our platform.
“It is not a magic pill that will once and for all eradicate every potential fraudster from a poker room. However, the sophistication of these tools is constantly improving, and they are continuously learning, resulting in more precise, selective, and optimal decision-making.”
Getting player engagement right
While igaming operators use ML for player engagement and moderation, they are presented with a risk that arises from system decisions related to player behaviour and rule enforcement. ML algorithms can process large amounts of data to detect irregular player behaviour. However, relying solely on ML technology without human supervision and interpretation can misread player actions and fail to detect cheating, collusion, game and bonus abuse.
Starostenkov adds, “A less obvious risk relates to an element of uncertainty about the level of decisions with which we can put our trust in AI. When employed for player moderation, support, or other operations, there is always a risk that we are over-reliant on AI.
“The success with which artificial intelligence tools complete simple tasks elevates this risk even further. Soon, in addition to AI trainers, AI strategists will also be a pivotal and in-demand role, specifically to calculate the risks of its involvement in our everyday activities.”
The risk increases further as igaming operators rely on historical data to train ML systems. If not properly calibrated and audited, algorithms could disproportionately flag and restrict certain player groups without adapting to evolving regulatory, tech and gaming trends.
The white-label challenge
EvenBet Gaming is a leading provider of poker solutions. The company supplies and supports platforms for two large poker networks, allowing players from various sites to play against each other in the same poker cash games and tournaments. By using a centrally-managed solution that utilises a shared infrastructure and resources, EvenBet Gaming provides operators with a quicker time to market, large player liquidity, easier poker room management, lower operational costs, server and hosting, compliance, tournament management and more.
Starostenkov says that using AI technology for platform and tool localisation is a game-changer. “White-label solutions are often difficult to adapt to specific markets, for example, in Asia or Africa. By powering them with AI, we can solve the most crucial problem of content localisation, not just through language adaptation, but also by tailoring the game offering itself to align with player demands.”
However, besides the benefits AI and ML bring, using these technologies in a white-label environment also creates risks. The nature of a centrally managed service often leads operators to rely on the service provider and not closely monitor security measures, fraud detection and player engagement protocols.
With a combined model of AI and human support to both operators and end customers, EvenBet Gaming is able to mitigate these risks.
Using AI in the second line of support
There is a vast difference between the frontline of customer support that operators provide players (B2C) and the support that a solution provider like EvenBet Gaming delivers to its business partners (B2B).
Starostenkov explains that the biggest contrast between existing practices in B2B and B2C operations lies in customer support. “For B2C operators, it is already normal practice to use AI in the first line of communication with players when they require quick responses to simple questions: how a bonus works, withdrawal timescales, solving login issues, and so on. In these situations, the machine is faster, more efficient, and takes an enormous communication load off company specialists, allowing them to focus on more intricate problems.”
“Our customer support is always employed at the point an operator is unable to solve a player’s problem or finds a notable technical issue. No casino operation is simple and standard, and that’s why we utilise the assistance of AI, to work with our knowledge database rather than with our customers.”
Embracing AI and ML
EvenBet Gaming is always looking to push the boundaries of technology. In 2012, it launched a native poker app, and a year later, it had already reached one million players on its platform. In 2014, EvenBet was one of the leaders in the igaming industry in implementing blockchain technologies and cryptocurrencies into its platform.
“We can try to predict how new AI models operate and learn compared to the existing ones, but even an educated guess is just a guess. There are four major categories with dozens of different algorithms, and new ones are continuously being created,” Starostenkov explains.
In 2024, the provider served more than 48 million users in 41 countries. So, like other educated guesses about technology that Starostenkov made throughout his tenure in EvenBet Gaming, it’s very likely that his working assumptions on AI and ML are also very accurate.