Raphie AI

As AI becomes increasingly embedded in iGaming customer service, the challenge is shifting from simply answering players to resolving issues accurately within the operator’s systems, policies and regulatory framework. Justin Heath, Co-Founder of Raphie, explains for iGaming Expert why effective AI support needs to combine live context, automation and human oversight.

AI has permeated customer service in every industry. In iGaming, however, fluency is only part of the requirement. Resolving a player issue often means combining live account information with operator policy, market-specific rules and the right escalation path. The question for operators is therefore not simply whether AI can hold a conversation, but whether it can work accurately inside the operating environment.

Conduet was founded in 2019 with a clear focus on delivering best-in-class player support for sportsbook and iGaming operators. From the outset, the view was that customer service should be more than an operational cost: done well, it can differentiate an operator, strengthen the player relationship and contribute directly to commercial performance. Building and running those teams gave us a detailed understanding of what good resolution requires in practice, from navigating operator systems and policies to taking the right action for the player.

That same focus was then applied to the development of our AI support platform, formerly gameLM, with the aim of increasing automation while limiting the risk that comes with it. Proprietary technology and years of operating experience have shaped how the platform performs in live gaming environments. From August 10, Conduet and gameLM become Raphie, one name for the business and the platform, and a clearer identity for the growth ahead.

What resolution requires in iGaming support

Player contacts often appear simpler than the work required to resolve them:

• Missing bonus: resolution may depend on eligibility, current account information and the terms of a specific offer.
• Pending withdrawal: can involve payment status, verification requirements and an approved next step.
• Responsible gaming signal: may require immediate escalation rather than an automated response.

These are different situations, but they place similar demands on the technology. It needs enough context to understand what has happened, determine the appropriate action and recognize when human judgment is required.

Raphie has been developed on a foundation of more than one million real iGaming interactions, alongside teams that work in player support every day. That operating base informs how the platform is trained, tested and measured. In live production, the AI Agents currently resolve roughly 74% of contacts, with player satisfaction around 88%.

The emphasis on resolution is deliberate because a conversation can be short, fluent and technically “closed” while leaving the underlying problem untouched. For operators, particularly in moments involving payments, account access or responsible gaming, that distinction has direct implications for service quality, operational workload and player trust.

Context, action and control

For AI to resolve rather than merely respond, it needs access to the systems where the work happens. Raphie connects with existing PAM, CRM and ticketing environments, using live player context and operator-approved rules to support defined actions within those workflows.

The platform is model-agnostic, allowing the underlying models to improve as the technology develops without requiring the surrounding operating process to be rebuilt. Guardrails, responsible gaming logic and human escalation sit around the AI, so cases that require judgment, sensitivity or additional review can move to trained people, which is a key requirement from operators, particularly at iGaming enterprise level.

Human oversight remains an important part of that control. Where the platform cannot establish the appropriate action with sufficient confidence, the case can be escalated to a trained human rather than forcing an automated outcome. In practice, that makes escalation part of the operating model, not a fallback.

A practical route to adoption

The same discipline applies to implementation. Raphie adapts to the operator’s existing tech stack. A typical proof of concept focuses on one defined use case over ten weeks, giving the operator a controlled way to evaluate performance on its own workflows before deciding where to extend the approach.

Security and governance form part of that operating model. Raphie is supported by SOC 2 Type II and GDPR-aligned controls, including encrypted data flows, PII redaction and auditability.

The move to Raphie reflects how the business has developed since starting in player support in 2019. The opportunity ahead is to apply that expertise more broadly as our iGaming AI support platform continues to grow, while maintaining the same focus on accuracy, control and measurable outcomes for operators and players.