Betsson France: Production AI for Regulated iGaming Support
Listen to this case study
0:00 / 0:00
75%
Customer Issues Resolved End-to-End
3.3 min
Average Resolution Time
40-50%
Cost Reduction
● Verified production results
Executive Summary
Betsson France operates one of the most demanding customer support environments in European iGaming: high-volume, French-licensed, governed by strict KYC, AML, and responsible gaming requirements, with sharp seasonality tied to the sporting calendar. Before InteractiveAI, support agents resolved tickets by manually navigating five siloed systems, and institutional knowledge lived in static documentation and individual managers' heads.
InteractiveAI deployed production-grade Interactive Agents into Betsson France's customer support operation, integrated end-to-end with PAM, Zendesk, Sumsub, Jira, and the promotion engine. The AI agents now run across Email, Live Chat, and KYC as a cross-functional layer interacting with Customer Support, Compliance, Fraud, and Operations.
~75% of customer issues are resolved by AI end-to-end. This combines 65-80% direct autonomous resolution on tickets that enter the queue with the 30+% of would-be tickets the AI eliminates proactively at root cause. KYC issues that used to take days now resolve in minutes. Every AI decision links back to a specific policy.
With InteractiveAI, we didn't deploy a chatbot. We built a governed AI system embedded in our operational stack. Roughly 75% of our customer issues are now resolved by AI end-to-end, plus 30+% of would-be tickets that never reach the queue at all, resolved proactively at root cause before a customer has to ask for help.
Mehdi Aounallah
Head of Customer Support, Betsson France
The Challenge
Betsson France ran a high-volume, high-regulation operation under structural friction. Within a single ticket, CS agents manually navigated five disconnected systems for player data, KYC verification, escalations, and bonus eligibility.
The cost per ticket was high. During peak sporting events, response times lengthened. Quality varied across shifts and agents. They needed an AI infrastructure layer that could take action under regulatory enforcement, at any hour, at any volume.
Betsson France evaluated standard helpdesk AI add-ons, in-house agent frameworks, and direct frontier model deployments before selecting InteractiveAI. The decision criterion that mattered: production AI they could trust under strict regulatory constraints, operate themselves, and extend across the rest of the business.
Business Impact
What surprised me most is how much of the AI we now run ourselves. When something needs adjusting, a new rule, a tone change for a sensitive topic, a tightening of a bonus eligibility check, I do it directly. I write the policy, test it, deploy it. No engineering ticket, no waiting cycle.
Mehdi Aounallah
Head of Customer Support, Betsson France
Beyond the Cost Line
What scales here is not chatbot deflection or FAQ retrieval but production support operations executed by agents that can be trusted to enforce policy precisely on every interaction. The business impact is structural: handling thousands of tickets a month now requires computing power, not extra headcount. When big sporting events cause sudden spikes in traffic, the AI absorbs the extra work instantly without the need to overstaff or slow down.
But beyond saving money and reducing workloads, this is a major upgrade in customer service. Faster response times and 24/7 availability directly improve the player's experience. Faster identity checks mean quicker first bets, faster withdrawals keep players loyal, and round-the-clock support means players get help instantly instead of waiting in a queue.
And beyond the numbers, the quality of service has changed: KYC issues that used to take days resolve in minutes, 24/7. The results exceeded every expectation we had.
Mehdi Aounallah
Head of Customer Support, Betsson France
Proactive Resolution
Most AI deployments measure success by deflection. Interactive Agents go further: they eliminate the conditions that produce tickets in the first place. The KYC agent operates directly on top of Sumsub, catching name-matching failures, document-quality issues, and verification edge cases that would previously surface as customer support tickets. The customer often never sees the failure state. The agent re-runs the check, completes the verification, and the player advances without writing in.
Conservative measured impact: 30+% of would-be tickets eliminated before reaching support. What used to be ~11,000 monthly tickets is now ~8,000, and the ~3,000 difference never enters the queue at all. This is the single highest-leverage outcome of the deployment: not just handling tickets faster, but reducing the volume of work the operation needs to handle.
“Thank you so much, I did not expect you to be able to resolve this issue so quickly and efficiently at this time of night.”
What the AI Handles
The largest domain. The agent resolves bonus eligibility, wagering questions, and welcome-offer disputes by reading the player's transaction history and account flags directly. Routine cases close in a single turn. Complex eligibility decisions, where commercial discretion is required, route to the bonus team with full context attached.
Identity verification is the most regulated workflow in iGaming and the highest-stakes test of any AI deployment. The agent reconciles verification data with the player profile, applies the verification policy at every level (basic ID, proof of address, IBAN), and replies in seconds with a full audit trail. Level 1 (ID): 85% autonomous, the agent performs fuzzy name matching and modifies Sumsub records to resolve mismatches, mimicking human reviewer behavior. Level 2 (POA): approaching 100% autonomous. During implementation, InteractiveAI also closed a long-standing gap: the previous system rejected French names with accents and minor address variations, generating unnecessary KYC failures the agent now resolves correctly.
Open-ended account questions and conversational support. This is where Interactive Agents differ from a traditional chatbot: the same agent that runs a deterministic compliance routine can hold a fluid conversation, and switch back the moment a regulated topic surfaces.
Withdrawal status questions and payment-pending diagnostics. The agent pulls withdrawal and ticket history, checks compliance flags, and either resolves directly or hands off to the payments team with a complete brief. Bank-account changes (where financial-fraud risk is highest) are routed to human review by design, not because the AI cannot handle them.
Personal-data, limit, and preference changes. The agent applies the right identity-verification gate, collects supporting documents, and either completes the update directly (low-risk changes like email) or hands off to operations with a complete brief (higher-risk changes like banking details).
A complex routine that touches multiple tools and teams, yet the agent resolves nearly all cases without escalation, a demonstration that autonomy on intricate workflows is achievable end-to-end.
Sportsbook history questions and deposit troubleshooting. Each follows the same pattern: account-state diagnosis, history retrieval, and structured escalation where human judgment is required.
Compliance & Governance
Every decision is traceable, testable, explainable, and auditable.
Mehdi Aounallah
Head of Customer Support, Betsson France
How It Works
Interactive Agents are composed from eight components that domain experts and engineers build together. Dynamic Context Engineering then retrieves only the policies that match the current customer interaction, so per-turn cost stays flat whether the policy library has 50 rules or 5,000.
Models are a commodity. Business impact comes from how context, data, and tools are managed across the entire AI lifecycle, from build through governance to continuous improvement. The agent that ships on day one is the weakest version of the agent that will ever run.
The Components
Persistent identity and guardrails.
Condition-action rules in natural language.
Multi-step procedures.
Domain vocabulary.
Reusable compositions of routines and policies.
Player state and account flags.
Gated connectors: PAM, Sumsub, Zendesk, Jira, IBAN/RIB, and InteractiveAI OCR.
The documented policy the agent grounds its answers in.
Managers write the rules. Engineering wires the systems. Both work in the same artifacts.
The Platform
Every decision captured with cost-per-run, latency, and compliance reports on demand.
Compliance edits policies and redeploys instantly, no engineering.
Agents, data, and 200+ models on cloud, private cloud, or on-premise, with regional residency and zero retention.
Every interaction becomes better future performance, validated in sandboxed experiments before production.
Platform Integrations
Source of truth for player data: account status, balances, personal details, limits, history. Read and write access enables the agent to retrieve player state and execute updates directly.
Email and Live Chat. The agent reads inbound tickets, classifies intent, executes workflows, writes responses, and manages full ticket lifecycle programmatically.
Identity verification: status, document review outcomes, KYC process triggers and follow-up.
Bank account validation for withdrawal eligibility checks.
Proprietary OCR for document-heavy workflows like bank account verification.
Bonus eligibility, offer state, and wagering data sourced for the Promotions & Offers routine.
Internal escalation routing for payment processing issues, compliance edge cases, fraud flags.
What Scales Next
Customer support was the first deployment. It is not the last. The AI infrastructure investment is one-time, not per-use-case: every additional Interactive Agent runs on the same platform, with policy-authoring skill, governance practice, and the datasets compounding from support traffic transferring to whatever comes next. The foundation is production-proven. What scales next is the business.
Take the Next Step
Book a 30-minute live demo where the InteractiveAI team walks through actual policies, routines, and traces from a working iGaming deployment.
InteractiveAI can introduce you directly to the Betsson France operations team.
A 6-week implementation on one or two of your highest-volume support domains, with measurable success criteria defined upfront.