Can AI really take over 70 per cent of gambling support?
AI could cut gambling operators’ customer support costs by as much as 75 per cent, according to a new whitepaper that argues most routine player queries can now be handled without a human agent.
The report, from iGaming support platform Tugi Tark, says operators using AI for around 70 per cent of support tickets could reduce the average cost of an interaction from about €1.88 in a human-only model to roughly €0.67 when AI and human agents work together.
The company prices its own AI-handled tickets at €0.15. In the report’s Romanian deployment example, a human-only support model costs about €1.88 per ticket.
The report says modern AI systems built specifically for gambling can handle “between 60 per cent and 80 per cent of player queries autonomously”, mostly in areas such as payments, account access, password resets, bonus rules and routine KYC guidance. Payment and fraud-related tickets alone account for 52 per cent of typical iGaming support volume, according to the whitepaper.
“In 2026, the ‘wait-and-see’ approach to AI is costing operators millions in unnecessary overhead,” commented Harpo Lilja, founder and CEO of Tugi Tark.
From cost centre to retention tool
The case is not only about cutting agents. The report frames AI support as a revenue-protection function, especially where withdrawals, failed deposits or blocked accounts are involved.
The whitepaper models a mid-sized operator with 100,000 monthly active players and a 5 per cent monthly churn rate, meaning that around 5,000 people stop playing each month. A 0.5 percentage point reduction in churn, it argues, would retain 500 additional players a month. At a conservative player lifetime value of €400, that would mean €200,000 in annual revenue protected.
According to the paper, it is the speed advantage enabled by AI that could make this possible. Human agents are benchmarked against a live-chat first response time of under 60 seconds. AI systems, the report says, can answer in under five seconds on chat and around three minutes by email.
Support is often triggered at moments of stress: a missing withdrawal, a failed deposit, a bonus dispute or an account restriction. A player waiting in a queue while money is involved is not merely inconvenienced: they may be one click away from a rival brand.
The regulatory brake
There is, however, a limit to the automation story. Gambling support sits inside a licensing environment that includes responsible gambling, AML, KYC and data-protection obligations.
The UK Gambling Commission’s current customer interaction guidance, for instance, requires remote licensees to monitor customer activity and identify harm from the point an account is opened. It also says operators must use indicators, including spend, patterns of spend, time spent gambling, customer-led contact and use of gambling management tools. The guidance adds that “staff know how to escalate a situation if they are unsure or require support”.
That point is crucial for AI. The Commission also says strong indicators of harm must be acted on in a timely way through automated processes, but where automated decisions significantly affect a customer, the customer must be told and allowed to contest them, with a manual review if they do.
Tugi Tark’s own report reflects that boundary. It says responsible gambling interactions should be routed rapidly to trained human agents, with AI providing context rather than resolving the underlying concern.
Regulators are likely to ask operators to open AI’s black box, industry commentators have warned. Paula Murphy, Head of Commercial at Mindway AI, a player-protection specialist whose tools use AI and neuroscience to detect at-risk gambling behaviour, made that point in a previous industry discussion on AI and gambling. “AI decisions must be explainable and understandable,” she said, adding that “black-box systems, where we can’t understand AI’s decisions, are unacceptable.”
A smaller support team, not a human-free one
The most likely near-term outcome is not a fully automated support floor. It is a smaller one.
Tugi Tark’s model suggests that at 70 per cent AI containment, a 17-agent team handling 18,000 tickets a month may need only five or six agents for escalated cases. Those remaining agents would deal with complex disputes, high-value players, responsible gambling cases and exceptions where human judgment is required.
That is a material workforce shift. For CFOs, the 70 per cent figure is attractive. The commercial prize is obvious: lower costs, instant replies and support that scales through sporting peaks without emergency hiring. The risk is just as clear: if AI closes tickets that should have been escalated, the savings may be short-lived.
Stay ahead with the stories driving global iGaming. Select SiGMA World as your source preference for exclusive voices, betting trends, prediction markets and key regulatory updates.
