Sportsbet, Flutter’s market-leading Australian brand, has tripled the number of customers its relationship managers (RMs) can handle, with no drop in satisfaction scores. This jump, from 400 to 1200, is no small feat in a business where personal service drives loyalty, and has been enabled by Jeeves, an artificial intelligence (AI) tool developed specifically for Sportsbet’s RM team. Like the fictional butler from whom it takes its name, Jeeves is designed to work invisibly in the background, keeping the RMs it supports operating seamlessly and efficiently.
“We know customers like having a dedicated relationship manager,” says Jenny Wishart, Head of Customer Platforms and Enablement at Sportsbet. “The question was, how can we make managing 3,000 customers feel like managing 300?” That’s the scale Sportsbet is testing toward. As part of a pilot to see how far the model can stretch, one RM is now trialing a portfolio of 2,500 customers.
Jeeves runs on its own standalone data platform, separate from Salesforce, the customer-relationship management (CRM) system that handles Sportsbet’s customer communications. The two are integrated, so Jeeves can read customer history and surface its recommendations directly inside the RM’s existing Salesforce workflow. It can understand a customer’s personal preference and know when a situation calls for a direct conversation, rather than a more templated response, allowing RMs to focus on the skilled, judgment-led customer interactions that help build loyalty.
The highest volume of inbound messages to Sportsbet’s RMs relates to customer generosity requests – with 42% of customer inquiries about bonus bets and other offers. The speed and quality of the response directly affect customer satisfaction. But customer language is informal and varied in ways that are challenging for automation. Rather than “Can I have my weekly bonus bet?”, a customer might ask: “Have you got anything for me today?”
Jeeves learned to interpret these messages using 12 months of real customer conversations. Now, when an inbound message arrives, Jeeves analyses it in seconds. It determines intent, checks whether the message is a generosity request, and if so, makes a recommendation: approve, decline or offer a smaller bonus than requested. It then drafts three personalized responses with slightly different tones, for the RM to review before sending. If Jeeves cannot determine intent, the request passes straight to the RM.
A key measure of how Jeeves is performing is how often RMs agree with its recommendations. A year ago, the rate was around 30%. By April 2026, it had reached 69%, driven by both broader adoption of the tool and a recalibration in January that aligned Jeeves’s decisioning more closely with customer sensitivity and commercial metrics. Beyond drafting responses, Jeeves can automatically apply an approved bonus to the customer’s account, removing another manual step.
Bet settlement queries are the second-highest volume inbound category and the next use case in development. Jeeves is being trained on three specific scenarios: T&C explanations, bet deductions, and late bet cancellations – where a betting market remains open after an event has started, allowing bets to be placed on potentially known outcomes. In these cases, Sportsbet’s Risk team cancels all bets placed after the official start time, which can result in expected winnings being revoked. These are among the most admin-intensive interactions Sportsbet handles, often involving several internal teams, and Jeeves is being designed to deliver consistent responses across all three situations.
“If you were just a user, you’d think Salesforce was making all these decisions,” Wishart says. The whole experience is seamless: Jeeves appears as a window in Salesforce, reducing a process that once required multiple systems to a couple of clicks. But no recommendation goes to a customer without being seen by an RM – they accept, adjust or override with a different response if needed before anything is sent.
“The question was, how can we make managing 3,000 customers feel like managing 300?” 
Designed with humans in the loop
Sportsbet’s approach to automation is deliberately cautious, not because the technology cannot go further, but because the company wants to protect human judgment. Jeeves leads on handling high-volume, routine decisions so RMs have more headspace and free time for nuanced engagements the algorithm cannot make.
Jeeves is deliberately designed to support responsible gaming by operating within strict scope limits and applying eligibility and safer-gambling checks, ensuring automated decisions are made only where it is appropriate and compliant to do so.
The current trial comes with clear guidelines: trust Jeeves for routine decisions, use its suggested messaging and intervene only when something looks wrong. Early signs are encouraging: the agreement rate, already at 69% in April, has climbed further still, to 72%. “It’s now about getting one person through that process so we can show the rest of the team that this works,” Wishart says.
The metric Sportsbet watches most closely is not whether customer satisfaction improves, but whether it holds. So far it has, even as customer portfolio sizes have tripled. If the 2,500-customer trial delivers the same result, it will further validate the model.
In parallel, a growing suite of outbound use cases are feeding RMs timely, insight-driven reasons to contact customers, complete with suggested language. “We can see if their activity has changed, or if there’s an insight we can act on,” says Wishart. “It gives the team a reason to check in and engage proactively with customers, with the right language and tone ready to go.”
Other Flutter brands are taking note. Sportsbet has shared a Jeeves overview with counterparts at FanDuel and UK & Ireland, with each brand charting its own path from there. FanDuel is pursuing its own use cases built on Salesforce Agentforce, while Sportsbet continues to expand Jeeves into new scenarios and is exploring how the platform could integrate with Agentforce to strengthen end-to-end capability, part of its ongoing work to scale AI-supported service without losing the human judgment at its core.
The original Jeeves made his master smarter and more capable without ever drawing attention to himself. Sportsbet’s version plays a similar role – ensuring that each RM can provide personalized service to vastly more customers thanks to the invisible magic of AI.
