Xpert Assistant in XpertOS: VIP escalation and bonus abuse exclusion in iGaming

Updated October 2, 2026

TL;DR: Sportsbook and casino operators lose VIP value when batch syncs push rewards past the end of a session, and lose bonus budget when flagged accounts keep receiving offers. Xtremepush's XpertOS helps customer relationship management (CRM) teams draft VIP rewards and bonus exclusions, with human sign-off before anything goes live. Xtremepush's governed data layer applies consent, suppression, self-exclusion, and responsible gambling rules outside the AI, so no agent can override them.

During a live match, a CRM manager at a UK sportsbook is watching a high-value player who has just hit a loyalty milestone. In another tab, the operator's fraud tool has flagged a cluster of new accounts sharing a device. This walkthrough follows both decisions through XpertOS, the agentic CRM operating system (OS) built into Xtremepush, and shows where people stay in charge.

Use case 1: How does a VIP reward land during a live match?

The milestone is worth most before the final whistle, while the player is still watching. When a standalone loyalty platform syncs player data on batch cycles, that reward lands the next morning, long after the session ended.

Xtremepush helps CRM teams spot emerging high-value players, nurture them towards VIP status, and flag those at risk of churning for a retention review. The operator's VIP team manages the relationship. XpertOS supports the CRM side of that handoff.

Which signals start a VIP escalation?

Casino and sportsbook CRM teams work with two kinds of VIP signal in Xtremepush. Real-time events come from PAM (player account management) backends via application programming interface (API) or Kafka, and from frontend software development kits (SDKs). A deposit or bet outcome is available in milliseconds and can trigger a journey. The milestone on the CRM manager's screen is one of these.

Scheduled scores come from Xtremepush's predictive models. Its churn model scores each active player's probability of churning and pushes results to segmentation daily or weekly, depending on the model.

Timing differs between the two. Milestones can trigger in-session, but churn scores refresh on a schedule, so churn-based escalations suit next-day outreach rather than same-session intervention.

How does Xpert Assistant speed up review?

Where VIP escalation depends on manual review, eligible players can wait for the next review cycle before they are recognised. Xpert Assistant shortens the list-building step. It is a natural language chat interface to CRM data, so the CRM manager can ask for a segment, such as high-value players with an open bet on the match in play, and get it back without writing a query. With Xtremepush's Business Intelligence connected, the assistant can also query event data, such as open bets, so segments draw on betting activity as well as player profiles.

Xpert Assistant then drafts the campaign. Campaigns it builds stay in draft state and cannot be sent, scheduled, or launched by the assistant.

What should an approver check before signing off?

A high-value player only receives a retention campaign after someone signs it off. Xpert Flows adds human approval gates at points the team chooses. A gate pauses the flow and notifies a designated approver.

The approver sees the proposed audience and campaign, then approves, edits, or rejects it. Some context sits outside the CRM data a draft draws on, such as what VIP managers know from recent conversations with their players.

Before approving a retention campaign, the approver checks that its criteria exclude players showing safer gambling indicators. A drop in activity can mean a player is cutting back on purpose. The UK Gambling Commission (UKGC) lists changes to deposit limits and repeated use of time-outs among the indicators of harm operators must monitor. A reward sent to that player works against them.

High-value players carry extra rules. Since 31 October 2020, UKGC rules have required operators to establish that a customer's spending is affordable and sustainable before treating them as a VIP, and to run ongoing harm checks after that.

Xtremepush's governed data layer applies responsible gambling flags outside any AI decision. When the operator maps its affordability or customer interaction markers into Xtremepush as player data, those flags hold the player out of retention and bonus sends, whatever an agent drafts.

Xtremepush's message categories can cap how many messages a player receives per day, week, and month. Caps belong to the category each campaign uses, so the VIP team can give rewards their own category, separate from the one used for general promotions, with its own limits. In a multi-stage journey, a category's conditions apply only when the player enters, so a reward step later in the journey is not held back by them.

How fast does an approved reward reach the player?

A VIP player's reward lands as soon as they qualify, because the journey behind it is approved in advance, not for each player. The VIP team or CRM manager signs off the journey and its criteria once, such as the milestone that triggers it and the safer gambling markers that exclude a player. From then on, Xtremepush runs it automatically for every player who qualifies.

XpertOS drafts these journeys, and Xtremepush's real-time infrastructure runs them. When the player on the CRM manager's screen meets the criteria mid-match, the milestone event reaches Xtremepush in milliseconds. The reward then goes out on the player's consented channels, including push, email, SMS, in-app, and onsite messages.

The send infrastructure is proven at scale. During the 2022 World Cup, LiveScore delivered the Argentina win announcement to millions in less than 5 seconds. The campaign audit log then records each change to the campaign with the date, the action, its source, and the user who made it.

That splits the work in two. Pre-approved journeys handle the in-session moment, where speed matters most. XpertOS earns its place on the considered decisions, such as a retention campaign built from the latest churn scores, where judgement matters more than seconds.

Use case 2: How do flagged accounts stay out of bonus campaigns?

Multi-account bonus abuse, like the accounts flagged in the second tab, ranks among the fraud threats iGaming operators worry about most. In Sumsub's 2025 iGaming survey, 63.8% of operators named bonus abuse as one of the most dangerous fraud schemes facing the industry, alongside identity fraud and money laundering at 64.8% each.

Detection sits with the operator's PAM, KYC (know your customer), and fraud tools. Those tools link accounts through device fingerprinting and behavioural signals. XpertOS and the governed data layer handle what happens next in CRM: keeping flagged players out of bonus campaigns and keeping an auditable record of the change.

What happens in CRM once an account is flagged?

A flagged player should stop receiving bonus offers while the fraud team investigates. The fraud tool's flag on the accounts in the second tab reaches Xtremepush through the PAM integration.

Xpert Assistant can then build a segment of flagged players, and the CRM manager confirms it as an exclusion from bonus campaigns. For a player who should receive no communications at all, a project-level send rule overrides every other rule. The assistant cannot change project settings, so that change is made in the Xtremepush platform.

How does a fraud flag reach every brand?

Operators running several brands need a flag raised on one brand to reach bonus campaigns on all of them. Fraudsters move from brand to brand, campaign to campaign. Operators running several brands on one PAM can run each brand as a separate project inside one Xtremepush account.

A segment in one project can draw on segment results from other projects in the same company. When the operator's fraud tooling flags an account, that flag can feed a cross-brand segment that excludes flagged players from bonus campaigns on every brand, provided the operator sends the flag into Xtremepush as player data. Xtremepush enables this cross-project option on request.

Who decides whether a flag is right?

Excluding a legitimate player on a false fraud flag can cost a VIP relationship. Legitimate players in the same household are a known source of false positives in fraud detection, because family members or flatmates may legitimately share a device or Internet Protocol (IP) address.

Determining whether shared signals point to multi-accounting, collusion, or a genuine household is an account management decision for the operator's fraud team, not a CRM one. Blocking or closing an account is a PAM action taken by the fraud or compliance team. Xtremepush controls who receives communications, not who can play.

If a reviewer finds a campaign exclusion was wrong, they reverse it, and the change is recorded in the campaign audit log.

Where does XpertOS stop?

For sportsbook and casino operators, XpertOS stops at the proposal. Campaigns it drafts stay in draft until someone approves them. It cannot change project settings. Account decisions stay with the operator's fraud and compliance teams.

How are wrong drafts caught?

AI-drafted VIP campaigns and flagged-player segments will sometimes be wrong for a specific player. What matters is whether the workflow catches the error before it reaches that player. iGaming CRM teams check XpertOS at two points. Every draft is reviewed at an approval gate before it goes live, and the audit log records each change with the user who made it.

1. Auditing the data behind a VIP escalation: If an escalation rests on a drop in session frequency, the CRM manager can check the player's session history to see whether the drop is real or a data quality issue. Where an operator already runs an internal model, such as a generosity or bonus-spend formula, the Xtremepush account team works with that logic directly rather than replacing it. The team shows which data points drive the model so the operator can validate and adjust it.
2. Improving drafts through reviewer feedback: Reviewers can rate any Xpert Assistant response with like and dislike buttons, and that feedback goes to Xtremepush to improve answer quality. Edits and rejections at approval gates keep wrong drafts from going live.
3. Escalating complex bonus fraud cases: A fraud ring working across several brands with device spoofing and behavioural mimicry needs specialist investigation. That work sits with the operator's fraud team. XpertOS keeps flagged players out of bonus campaigns while the investigation runs, and does not replace the fraud team.

How does a VIP manager see why a player is at risk?

A VIP manager needs to know why a player appears on an at-risk list before acting on it. Xtremepush uses explainable AI to show the key drivers and accuracy of every predictive model.

Explainability matters to regulators as well. Guidance from the Information Commissioner's Office (ICO) expects organisations using AI-assisted decisions to be answerable to regulators and to the people affected, and to evidence how those decisions are made. Scoring that cannot be explained is a compliance risk.

How should approvers read churn scores?

Churn scores for high-value players are probabilities, not verdicts. A higher score means the model rates churn as more likely. Operators decide which score range puts a player into an escalation segment, and which safer gambling markers keep a player out of it.

What stops an AI sending a bonus to a self-excluded player?

A self-excluded player who receives a bonus offer is a regulatory breach, whatever a model predicted. Under UKGC rules, operators must take all reasonable steps to prevent any marketing material being sent to a self-excluded customer, and remove or flag that customer in their marketing databases within two days.

Xtremepush's governed data layer applies suppression lists, self-exclusion rules, consent, and jurisdictional logic independently of any agent. To keep offers inside licensed markets, an operator can set a project-level send rule on a player attribute it maps in, such as licensed market, and no communication goes to a player who fails it.

What can an agent never override?

Sportsbook and casino operators can now choose from CRM platforms that ship AI agents as standard. In regulated markets, full autonomy is the wrong goal. Compliance has to hold whatever an agent proposes, so it belongs in the data layer, outside any agent.

XpertOS is built around that belief. Its agents reason, plan, and propose actions inside set guardrails. They draw on Xtremepush's governed data layer, which holds the customer record, consent, policies, and regulatory logic. Xtremepush then sends only what that layer permits.

Approval gates are only part of the answer. XpertOS takes list-building and drafting off the CRM team, while people decide which campaigns run. VIP relationships stay with the VIP team. Fraud decisions stay with the fraud and compliance teams.

The compliance rules sit underneath all of it. Xtremepush's governed data layer applies consent, suppression, self-exclusion, and responsible gambling rules independently of any agent. No proposal from XpertOS can override them. A self-excluded player stays out of every marketing send, whatever an agent drafts.

Map your VIP and bonus exclusion workflows with our team to see where pre-approved journeys, Xpert Assistant, and the governed data layer fit in your own setup.

FAQs

How does AI handle VIP escalation without annoying high-value players?

Xtremepush's governed data layer applies consent checks, suppression rules, and, where message categories are set, per-player frequency caps. Every campaign Xpert Assistant drafts stays in draft until a VIP team member or CRM manager approves it, so they can reject or edit anything that risks over-messaging a VIP.

Can XpertOS keep flagged bonus abusers out of campaigns across multiple brands?

Yes, once cross-project segments are enabled for the account and the operator's PAM or fraud tool sends the flag into Xtremepush as player data. A flag raised on one brand can then exclude the player from bonus campaigns on the others, with each brand running as a separate project.

What happens if a legitimate VIP is flagged as a fraud risk?

A false flag should not cost the relationship, and in Xtremepush it only pauses bonus offers. Xtremepush does not touch the account, and the player still receives VIP messages outside bonus campaigns. The fraud or compliance team reviews the flag and owns any account decision. If the flag was wrong, the exclusion is reversed, and the campaign audit log records the change.

Does this replace my CRM team or augment them?

XpertOS augments the CRM team. Xpert Assistant finds segments and drafts campaigns, while approval gates keep people in charge of every consequential step. XpertOS is designed to deliver 10x campaign output from the same team size. It does not replace CRM managers, VIP managers, or fraud specialists.

How long does it take to go live?

Xtremepush's average implementation time is 6 to 8 weeks, covering technical integration and strategic account setup. Fraud flags and VIP attributes reach XpertOS once the operator maps them in from its PAM backend. Churn scores come from Xtremepush's own predictive models. Every operator gets a dedicated account manager with 8 hours per month of strategic support, covering data setup, platform training, and market-specific practice drawn from 250+ operators.

Key terms glossary

Agentic AI: AI that can reason, plan, and propose actions inside set guardrails. XpertOS agents propose actions, while Xtremepush's governed data layer enforces compliance independently.

VIP escalation: Identifying high-value players who show a signal that needs attention, such as a milestone or churn risk, and routing a proposed campaign to the operator's VIP team or CRM manager for approval. Once approved, it runs for every player who meets its criteria.

Bonus abuse: The systematic exploitation of promotional offers to extract value without genuine play, often through multi-accounting, device spoofing, or collusion.

Multi-accounting: One person opening several accounts, often across an operator's brands, to claim the same promotion more than once. Fraud tools link the accounts through shared device and behavioural signals.

Collusion: Separate players acting together to exploit a promotion or game, for example by coordinating their bets. Each account belongs to a different real person, which is what separates collusion from multi-accounting.

Human-in-the-loop: A design pattern where an AI system proposes an action but a human must approve, edit, or reject it before execution. In XpertOS, actions that change things, such as creating a campaign, pause for approval.