Why iGaming AI Support Metrics Lack Industry Comparability

Metric Definitions Drive Divergent Figures

A single deployment of AI player support software can report an 85% containment rate, a 70% automation rate, and a 55% true resolution rate, depending on the specific metric quoted. According to an analysis published by European Gaming on October 22, 2026, these figures are frequently conflated by operators and procurement teams, leading to misleading comparisons between vendors.

The author of the analysis, who disclosed running a company that sells AI player support, identified three distinct metrics that are often treated as interchangeable: containment, automation, and resolution. Containment counts conversations that never reached a human agent, including cases where a player gives up and closes the chat. Automation counts conversations where the AI produced every reply, regardless of whether the underlying problem was fixed. True resolution is defined as the issue being fixed and the player not returning within a set window, typically three to seven days.

“A player who gives up and closes the chat counts as contained,” the author noted. “None of that is dishonest in itself. But a figure without its denominator cannot be compared with another figure without its denominator.”

Denominator Issues and Deployment Context

Beyond definitions, the scope of data included in these figures varies significantly. Some reported metrics cover only chat channels, excluding email where more complex cases often land. Other figures exclude tickets routed directly to payments or compliance teams. Additionally, many vendors cite performance from the initial deployment period, often the first eight weeks, when the AI handles a hand-picked set of simple intents.

In regulated gambling markets, the stakes of inaccurate resolution reporting are higher than in other sectors. Critical tickets involve withdrawals, identity verification, bonus disputes, and responsible gambling signals. The UK Gambling Commission’s complaints data identifies withdrawals and financial transactions as the largest single category of complaints. If an AI system contains a conversation about a withdrawal without actually resolving the issue, the player may return angrier, potentially leading to a formal regulatory complaint.

“In e-commerce, a falsely ‘resolved’ ticket usually means a slightly annoyed customer,” the author explained. “A player told that their withdrawal is ‘being processed’ when it is actually blocked on a source-of-funds check will come back, angrier, and in some markets that becomes a complaint to the regulator.”

Forward-Deployed Models and Vendor Performance

Automation rates are not solely determined by the software itself but also by the vendor’s post-contract support. This includes mapping bonus terms, KYC rules, payment flows, and escalation paths. Vendors that utilize a “forward-deployed model”—sending engineers and iGaming specialists to work inside the operator’s operation for the first weeks—tend to achieve higher resolution rates faster than those that simply provide a login and a knowledge base upload.

Domain expertise, such as understanding wagering requirements and responsible gambling signals, also impacts both resolution rates and safety. Furthermore, operator commitment plays a role; giving AI back-office access and dedicating support lead time to review outcomes affects final performance. Consequently, the same software can produce different automation rates at two different operators based on these environmental factors.

Call for a Shared Industry Standard

The author advocates for a shared public definition of AI resolution that includes a fixed re-contact window. The proposed standard should also include per-category reporting for payments, KYC, bonuses, and responsible gambling. Trade bodies and larger platforms are identified as the entities best placed to establish this standard.

To assist operators in evaluating vendors, the analysis recommends asking five specific questions: the definition of the number, the denominator channels included, the measurement period, specific performance on withdrawals and KYC, and access to sample conversations. “A vendor that is confident in its figure should be able to answer all five in writing,” the author stated. “Until then, the safest assumption for any operator is that two percentages from two vendors are measuring two different things.”