BetConstruct AI executive outlines personalisation, AI agents and fraud detection for iGaming operators

Mushegh Khachatryan, Chief AI Officer at BetConstruct AI, has described in an SBC News interview how the company uses artificial intelligence to help iGaming operators cut costs and raise revenue. He said personalisation goes beyond game recommendations and covers the casino lobby, promotions, messaging and even decisions about when not to contact a player. The stated aim is to move from one casino experience for a million players to a million tailored experiences.

On commercial results, Khachatryan said there is no universal figure because outcomes depend on the operator, market and product maturity. Across the company's deployments, recommendation systems have delivered uplifts from high single digits to significantly higher levels, and more mature implementations can improve turnover by around 10-15% in the areas where personalisation is applied. He said the gains come mainly from reducing friction, helping players find relevant content faster, which can lift session depth, game discovery, bet frequency and retention. He stressed that the objective is relevance rather than simply maximising playing time.

Khachatryan also discussed AI agents, which he said connect existing recommendation, churn and fraud models with operational tools. Under this approach, an operator could ask in plain language what happened to casino revenue, which player segments drove a decline, and then request a campaign for the segment with the highest churn risk. BetConstruct AI is building its Operational Copilot around this conversational interface, moving from dashboards that display data towards systems that understand, analyse and eventually help execute tasks.

He said agents work best where there is substantial data, repetition and measurable outcomes, such as CRM, analytics, player segmentation, campaign preparation, customer support and operational monitoring. Humans, he argued, must remain responsible for significant financial decisions, responsible gambling interventions, regulatory decisions and unusual customer situations.

On risk, Khachatryan said machine learning can analyse hundreds of behavioural signals at once, including deposits, withdrawals, betting patterns, bonus usage and game selection, to flag statistically unusual activity for review before a rule is broken. The same approach is applied to negative-NGR prediction, hidden VIP detection and churn models.

Finally, he said the company's architecture is model-agnostic, allowing it to test new AI models against real iGaming use cases and adopt better ones quickly. In his view, the lasting advantage lies in an operator's data, domain knowledge and ability to turn models into useful products.