iGaming Platform AI & Automation Comparison
A field-level comparison of AI decisioning, prediction, risk, player protection, conversational tools, content generation and accountable ownership. Generic AI claims remain separate from explicit named-product assignments and every conclusion keeps its delivery boundary.
Last updated August 20, 2026
AI capability, not an AI score
The matrix preserves nine separate conclusions and does not add them into a feature score. Provider order follows the site's holistic editorial score. A Yes, Partial or named product does not prove that a model is native, real-time, production-deployed, included in every contract or effective against an agreed benchmark.
142/162
scoped conclusions
58/162
verified conclusions
30/162
supported conclusions
54/162
editorial inferences
8/18
providers with at least one named AI product scope; this is a scope count, not an AI-maturity score
Resolved includes Yes, No, Partial and other scoped conclusions. No is used only for a bounded negative conclusion; an unknown capability remains Unresolved rather than being converted into No.
Provider records
A named product appears only when an AI conclusion is attached to that offer. Its absence does not erase a provider-level conclusion. Every value below retains its description, confidence and product or deployment boundary.
01EveryMatrix8.8
Named AI product scope
- EngageSuite · AI dynamic bonusing
- Bonus Guardian · AI fraud detection
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
Future Anthem Real-Time Experiences and Content Recommendations are integrated across EveryMatrix brands.
- AI recommendation engine
Yes
Verified conclusion
Native game recommendations plus Future Anthem real-time content recommendations.
- Churn / LTV prediction
Yes
Supported conclusion
Real-time churn prevention is available; a standalone native LTV model specification remains unresolved.
- AI dynamic bonusing
Yes
Verified conclusion
Future Anthem's integrated Real-Time Experiences tailors bonus interventions to individual behavior.
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
Bonus Guardian uses adaptive behavioral ML against bonus abuse, multi-account rings, proxies, synthetic identities, and bots.
- AI responsible gambling
No
Editorial inference
Current AI products cover fraud, personalization, recommendations and churn; responsible-gambling controls exist, but no validated AI harm-detection product is established.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
EveryMatrix has no operator- or player-facing AI chatbot product.
- AI content generation
No
Editorial inference
EveryMatrix has no productized AI content-generation entitlement; generic AI terms do not establish a shipped product.
- Dedicated AI leadership
No
Editorial inference
EveryMatrix is recruiting a Chief AI Officer and an AI Lead; no group-level AI executive was appointed as of July 11, 2026.
02Playtech8.7
Named AI product scope
- PAM+ SaaS Ecosystem · AI chatbot
- BetBuddy within Playtech Protect · AI responsible gambling
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
- AI recommendation engine
Yes
Verified conclusion
Real-time player data, predictive segmentation and personalisation feed recommended content, journeys and offers through PAM+ and Engagement Centre.
- Churn / LTV prediction
Yes
Verified conclusion
- AI dynamic bonusing
Yes
Supported conclusion
Real-time decisioning, personalized player journeys and dynamic bonus structures are active capabilities.
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
- AI responsible gambling
Yes
Verified conclusion
Runs on BetBuddy, an AI behavioral-monitoring engine that risk-scores players and triggers safer-gambling nudges.
Conversational, generative and governance scope
- AI chatbot
Partial
Verified conclusion
Conversational AI is available through the integrated Ada SaaS partner rather than as a Playtech-built chatbot.
- AI content generation
Partial
Editorial inference
Managed Services includes AI avatars and chatbots; a general operator-facing generative game or campaign-content product is not established.
- Dedicated AI leadership
Yes
Verified conclusion
Playtech now has a Central AI Team and AI Centre of Excellence, an AI Committee, an AI Delivery & Operations Manager function and an AI Project, Governance & Operations Manager function. A named group CAIO is not required for this field to be Yes.
03SOFTSWISS8.3
Decisioning, prediction and player value
- AI personalization
Yes
Supported conclusion
Driven by the DOSSIER behavior service and real-time player segmentation feeding personalized campaigns and offers.
- AI recommendation engine
Partial
Supported conclusion
Sportsbook event ranking and DOSSIER personalisation are supported; a universal native casino-game recommender is not established.
- Churn / LTV prediction
Yes
Supported conclusion
- AI dynamic bonusing
Yes
Supported conclusion
ML-derived audiences can feed the Bonus API, including free-spin activation; a fully autonomous AI bonus-decision engine is not established.
Risk and player protection
- AI fraud detection
Yes
Supported conclusion
ML supports anomaly and traffic-quality scoring; final investigation and decisions remain with anti-fraud teams.
- AI responsible gambling
Yes
Supported conclusion
The ML Risk Scoring Tool identifies risky behaviour and prioritises responsible-gambling cases; model performance and outcome validation remain unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
- AI content generation
Partial
Verified conclusion
Its WebStudio uses AI to help artists generate creative concepts and visuals, not automated promo copy across the platform.
- Dedicated AI leadership
Yes
Verified conclusion
Appointed Denis Romanovskiy as its first Chief AI Officer in January 2026 to run an enterprise AI platform and strategy.
04Pragmatic Solutions8.2
Decisioning, prediction and player value
- AI personalization
Partial
Supported conclusion
Available through third-party CRM/CDP tools or operator models; no native Pragmatic AI engine was established.
- AI recommendation engine
Unresolved
No mature native recommendation engine was established.
- Churn / LTV prediction
Unresolved
Data Lake can feed operator or third-party models, but a production churn- or lifetime-value-prediction capability within the retained platform scope remains unresolved.
- AI dynamic bonusing
Yes
Supported conclusion
Partner-enabled through pre-integrated Smartico and AI-led CRM options: Smartico supports real-time dynamic bonus formulas, personalisation and automation. The bonus decisioning is not a native Pragmatic model and may be an add-on service.
Risk and player protection
- AI fraud detection
Partial
Supported conclusion
External KYC and fraud integrations can add models; core native model depth remains unresolved.
- AI responsible gambling
Partial
Supported conclusion
Partner tools such as Crucial Compliance provide behavioural models; these are not owned by Pragmatic.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
A native chatbot product is not established.
- AI content generation
No
Editorial inference
A native generative-content product is not established.
- Dedicated AI leadership
No
Editorial inference
A dedicated AI executive is not established; the Technical Architect (AI) role indicates capability under development.
05Aristocrat Interactive8.1
Named AI product scope
- Aristocrat Interactive iLottery · AI responsible gambling
- Gaming Analytics · Churn / LTV prediction
Decisioning, prediction and player value
- AI personalization
Partial
Supported conclusion
NeoEngage supports iLottery personalization and Gaming Analytics supports land-based player engagement; this is not one PAM-wide AI layer.
- AI recommendation engine
Partial
Supported conclusion
Gaming Analytics provides land-based slot recommendations; an online-casino game recommender is not established.
- Churn / LTV prediction
Yes
Supported conclusion
Gaming Analytics provides churn and future-value models for land-based casino player development.
- AI dynamic bonusing
No
Editorial inference
Bonus and AI capabilities exist in separate products, but no AI-driven dynamic bonusing system is established.
Risk and player protection
- AI fraud detection
Unresolved
Fraud monitoring is available, but AI-specific detection is not established in the retained PAM.
- AI responsible gambling
Yes
Supported conclusion
NeoSafe applies machine learning to iLottery safer-play risk monitoring; portfolio-wide equivalence is not established.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
Michigan includes customer-service chatbot support, but an AI chatbot capability is not established.
- AI content generation
No
Editorial inference
No current production AI content-generation product is established.
- Dedicated AI leadership
Yes
Verified conclusion
Kiran Brahmandam became Managing Director, Gaming Analytics after the January 2026 acquisition.
06Altenar8.0
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
- AI recommendation engine
Yes
Verified conclusion
Bet Suggestions and Altenar Intelligence Services use aggregated betting behaviour to refresh recommendations and promotional content.
- Churn / LTV prediction
Yes
Editorial inference
Fast Track predictive player modelling is integrated with Altenar; native or default Altenar-module delivery is not established.
- AI dynamic bonusing
Yes
Supported conclusion
Altenar Intelligence Services can automatically populate Bet Cards and boost campaigns from behavioural and selection data.
Risk and player protection
- AI fraud detection
Partial
Editorial inference
AI is applied to fraud prevention, match-fixing detection and automated AML, alongside ongoing behavioural fraud and transaction monitoring in the PAM. The deployed model, coverage and validation metrics remain unresolved.
- AI responsible gambling
Partial
Editorial inference
Altenar uses reviewed and monitored ML models to improve responsible-gambling interventions. A production harm model, jurisdictional scope and validation metrics remain unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
A native player-facing or operator-support AI chatbot is not established. Customer-specific third-party integrations remain possible.
- AI content generation
Partial
Editorial inference
Available through current Fast Track AI capabilities when that integration and feature set are licensed; no native Altenar content generator is established.
- Dedicated AI leadership
No
Editorial inference
Data analysts and product leaders share ML ownership; a dedicated AI executive, head of AI or equivalent owner is not established.
07Kambi8.0
Decisioning, prediction and player value
- AI personalization
Yes
Editorial inference
AI personalisation of betting content/offers via Shape Games and network data.
- AI recommendation engine
Yes
Verified conclusion
Bet recommendations and same-game/bet-builder suggestions, driven by Kambi's data and AI trading.
- Churn / LTV prediction
No
Editorial inference
Kambi has AI trading, recommendation and personalisation capabilities, but no dedicated AI churn/LTV product is established.
- AI dynamic bonusing
No
Editorial inference
AI drives pricing and recommendations; dynamic bonusing is not an established Kambi product.
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
AI trading and risk flag arbitrage, bonus abuse and suspicious betting; a 24/7 Sportsbook Control team monitors integrity.
- AI responsible gambling
Partial
Verified conclusion
Kambi's Green Values approach and player profiling analyse behavioural changes, create risk profiles and alert operators; human analysts confirm material changes rather than relying solely on automated decisions.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
No Kambi AI support chatbot is established.
- AI content generation
No
Editorial inference
Kambi's production AI focus is pricing, trading, risk and recommendation, not generative marketing content.
- Dedicated AI leadership
Yes
Verified conclusion
Tzeract is a dedicated AI-driven trading division, a strong signal of AI seriousness.
08Light & Wonder7.7
Decisioning, prediction and player value
- AI personalization
No
Verified conclusion
No separate iGaming AI product is established; personalization is rules- and data-driven.
- AI recommendation engine
No
Verified conclusion
No separate AI recommendation engine is established.
- Churn / LTV prediction
Unresolved
No native churn or LTV model is established.
- AI dynamic bonusing
No
Verified conclusion
Rules- and segment-driven engagement rather than an established AI optimizer.
Risk and player protection
- AI fraud detection
Partial
Verified conclusion
Via integrated third-party risk providers, not an established L&W model.
- AI responsible gambling
No
Verified conclusion
No native responsible-gambling AI model is established; OPS provides player-safety data and controls.
Conversational, generative and governance scope
- AI chatbot
No
Verified conclusion
No core product.
- AI content generation
No
Verified conclusion
No core product.
- Dedicated AI leadership
Unresolved
No iGaming-specific AI leader is established.
09Pariplay7.5
Decisioning, prediction and player value
- AI personalization
Unresolved
A provider- or partner-delivered AI personalization capability within the retained Pariplay scope remains unresolved.
- AI recommendation engine
Unresolved
A provider- or partner-delivered AI recommendation engine within the retained Pariplay scope remains unresolved.
- Churn / LTV prediction
Unresolved
A provider- or partner-delivered churn- or lifetime-value-prediction capability within the retained Pariplay scope remains unresolved.
- AI dynamic bonusing
Unresolved
An AI bonus-decisioning capability within the retained Pariplay scope remains unresolved.
Risk and player protection
- AI fraud detection
Unresolved
An AI fraud-detection model within the retained Pariplay scope remains unresolved.
- AI responsible gambling
Unresolved
An AI or machine-learning safer-gambling model within the retained Pariplay scope remains unresolved.
Conversational, generative and governance scope
- AI chatbot
Unresolved
A player- or operator-facing AI chatbot within the retained Pariplay scope remains unresolved.
- AI content generation
Unresolved
An operator-facing AI content-generation capability within the retained Pariplay scope remains unresolved.
- Dedicated AI leadership
Unresolved
A dedicated Pariplay AI executive or accountable AI owner remains unresolved.
10Digitain7.1
Named AI product scope
- Centrivo CRM · AI recommendation engine
Decisioning, prediction and player value
- AI personalization
Yes
Supported conclusion
Digitain's AI engine recommends games, bonuses and features and dynamically personalizes content. Paydrom separately uses machine learning for behavior-based quick-deposit suggestions.
- AI recommendation engine
Yes
Supported conclusion
- Churn / LTV prediction
Yes
Supported conclusion
AI supports at-risk churn identification and proactive re-engagement. A separate LTV prediction model and validation metrics remain unresolved.
- AI dynamic bonusing
Yes
Supported conclusion
Digitain's AI Recommendation Engine recommends player-specific bonuses, while Centrivo CRM delivers targeted campaigns and dynamic bonus rules. Fully autonomous bonus issuance is not established.
Risk and player protection
- AI fraud detection
Partial
Editorial inference
AI and machine learning are used across the products, including AI-driven player profiling; Paydrom adds anti-fraud rules and suspicious-account detection. A production ML fraud model, training data and measured detection performance remain unresolved.
- AI responsible gambling
Partial
Supported conclusion
Behavioral monitoring, AI-powered risk detection and predictive interventions identify problematic play. Model performance remains unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
No native operator- or player-facing AI chatbot product is established.
- AI content generation
No
Editorial inference
AI use is limited to a marketing campaign rather than an operator-facing content-generation product.
- Dedicated AI leadership
No
Editorial inference
Digitain has a specialist AI and machine-learning team, but no dedicated AI executive or accountable AI-governance leader is established.
11White Hat Gaming7.1
Decisioning, prediction and player value
- AI personalization
Partial
Editorial inference
AI-labelled personalisation is available through CRM partners rather than a White Hat-owned model.
- AI recommendation engine
Partial
Editorial inference
Optimove can provide AI-ranked game and lobby recommendations over White Hat event data. The capability is partner-delivered and deployment-specific, not White Hat-native.
- Churn / LTV prediction
Yes
Supported conclusion
Predictive churn, future value, reactivation and VIP likelihood are available through the integrated Optimove stack, not White Hat-owned models.
- AI dynamic bonusing
Yes
Editorial inference
White Hat's bonus engine can be orchestrated with partner AI promotion decisioning. Bonus execution is White Hat; AI selection is partner-led and deployment-specific.
Risk and player protection
- AI fraud detection
Partial
Supported conclusion
Shufti supplies AI-driven document authentication and facial verification; White Hat also has payment-risk rules and alerts. A proprietary White Hat fraud-ML model is not established.
- AI responsible gambling
Unresolved
A proprietary behavioral algorithm uses more than 50 harm markers, but an AI or machine-learning basis remains unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Verified conclusion
An AI customer-support chatbot is not established.
- AI content generation
Partial
Editorial inference
Optimove includes generative campaign-content assistance; native or universally licensed PAM delivery is not established.
- Dedicated AI leadership
No
Verified conclusion
An executive or product role dedicated specifically to AI is not established.
12GR8 Tech7.0
Named AI product scope
- GR8 CRM and AI · AI recommendation engine
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
- AI recommendation engine
Yes
Verified conclusion
- Churn / LTV prediction
Yes
Verified conclusion
Churn prediction and LTV measurement are established; a separate LTV forecasting model is not.
- AI dynamic bonusing
Yes
Verified conclusion
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
- AI responsible gambling
Partial
Editorial inference
Behavioral risk signals, segmentation and automated restrictions can support safer-gambling interventions; a dedicated AI harm-detection model and outcome validation remain unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
GR8 uses LLMs for campaign text and translations; no player-support chatbot product is established.
- AI content generation
Yes
Verified conclusion
- Dedicated AI leadership
Yes
Verified conclusion
Raid Arfua is Head of Artificial Intelligence.
13GiG6.8
Named AI product scope
- DataX, LogicX & GiG Assistant · AI personalization · AI fraud detection · AI chatbot
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
DataX, LogicX and GRE 2.0 are proprietary; the newest deep-learning sports and casino personalization layer is supplied by VAIX, a Sportradar company.
- AI recommendation engine
Yes
Verified conclusion
- Churn / LTV prediction
Yes
Verified conclusion
- AI dynamic bonusing
Yes
Verified conclusion
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
- AI responsible gambling
Yes
Verified conclusion
LogicX and the Crucial Compliance integration combine behavioural analysis with automated AML/RG risk scoring and interventions.
Conversational, generative and governance scope
- AI chatbot
Partial
Verified conclusion
GiG Assistant is an operator-facing natural-language intelligence layer, not primarily a player customer-service bot.
- AI content generation
No
Editorial inference
GiG has no dedicated generative-content product in its current portfolio.
- Dedicated AI leadership
No
Editorial inference
No standalone AI executive is named; ownership sits across product, data and executive leadership.
14Bragg Gaming Group6.7
Decisioning, prediction and player value
- AI personalization
Partial
Verified conclusion
Current but scoped to Fuze behavioural intelligence and recommendations.
- AI recommendation engine
Yes
Verified conclusion
Yes — AI-powered game recommendation in Fuze.
- Churn / LTV prediction
Unresolved
Behavioural intelligence targets lower churn and higher lifetime value, but a churn- or lifetime-value-prediction model and its validation remain unresolved.
- AI dynamic bonusing
Yes
Verified conclusion
Fuze mechanics can adapt to player behaviour.
Risk and player protection
- AI fraud detection
Unresolved
Fraud rules and monitoring are established, but a production AI or machine-learning fraud-detection model remains unresolved.
- AI responsible gambling
Unresolved
Automated behavioural risk detection is established, but an AI or machine-learning basis remains unresolved.
Conversational, generative and governance scope
- AI chatbot
No
Verified conclusion
A current Bragg AI chatbot product is not established.
- AI content generation
No
Verified conclusion
A current customer-facing generative-content product is not established. AI-assisted internal workflows are part of the 2027 roadmap.
- Dedicated AI leadership
Yes
Verified conclusion
Luka Pataky, EVP of AI and Innovation; COO Morten Tonnesen also leads the AI-first transformation. Bragg AI Brain was initiated in 2026 with targets for 2027, so leadership commitment is ahead of delivered platform breadth.
15Slotegrator6.3
Decisioning, prediction and player value
- AI personalization
Partial
Editorial inference
Decision support and segmentation exist; model performance remains unresolved.
- AI recommendation engine
Unresolved
The ChatGPT-style BI assistant provides business suggestions, but a player-facing game recommendation engine remains unresolved.
- Churn / LTV prediction
No
Editorial inference
No shipped Slotegrator churn or lifetime-value model is established.
- AI dynamic bonusing
No
Editorial inference
No autonomous dynamic-bonus model is demonstrated.
Risk and player protection
- AI fraud detection
Unresolved
An AI assistant summarizes existing risk metrics, but an AI fraud-detection model remains unresolved; rules, IP and device signals are the established controls.
- AI responsible gambling
No
Editorial inference
Responsible-gambling guidance discusses monitoring, but no dedicated AI RG product is demonstrated.
Conversational, generative and governance scope
- AI chatbot
Partial
Editorial inference
A ChatGPT-style BI/business assistant exists; universal player support does not.
- AI content generation
No
Editorial inference
No established productized content-generation feature.
- Dedicated AI leadership
No
Editorial inference
A dedicated AI or data-science executive is not established.
16BetConstruct6.0
Named AI product scope
- BetConstruct AI · AI recommendation engine · AI fraud detection · AI personalization
Decisioning, prediction and player value
- AI personalization
Yes
Verified conclusion
- AI recommendation engine
Yes
Verified conclusion
- Churn / LTV prediction
Yes
Verified conclusion
- AI dynamic bonusing
No
Editorial inference
The bonus engine and AI segmentation can be combined, but a live AI system that autonomously sets bonus value is not established. Next Best Action remained a 2026 roadmap item.
Risk and player protection
- AI fraud detection
Yes
Verified conclusion
- AI responsible gambling
Yes
Verified conclusion
Umbrella combines configurable regulatory rules, transaction monitoring, real-time alerts and markers-of-harm detection for AML and responsible-gambling teams.
Conversational, generative and governance scope
- AI chatbot
Partial
Verified conclusion
Player support chat comes through the iGaming BOT's integration of the third-party Hoory AI assistant, not a chatbot BetConstruct builds itself. Its own AI Analyst is an operator query tool, not player support.
- AI content generation
No
Editorial inference
The AI suite predicts, segments, recommends and analyses; generative campaign-copy and creative production are not established capabilities.
- Dedicated AI leadership
Yes
Verified conclusion
BetConstruct AI has a dedicated CEO, Lena Yasir, appointed in June 2026, and the company is actively hiring around the AI-led product strategy.
17SoftGamings5.8
Decisioning, prediction and player value
- AI personalization
No
Editorial inference
Native AI personalisation is not established.
- AI recommendation engine
No
Editorial inference
A native AI recommender is not established.
- Churn / LTV prediction
No
Editorial inference
- AI dynamic bonusing
No
Editorial inference
No established AI decisioning; bonus rules are configurable.
Risk and player protection
- AI fraud detection
Partial
Supported conclusion
Partial through Sumsub and managed sportsbook risk. AI and machine learning are limited to partner identity and fraud tooling plus the sportsbook risk service, not a unified in-house casino fraud model.
- AI responsible gambling
No
Editorial inference
A native AI harm-detection product is not established.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
- AI content generation
No
Editorial inference
- Dedicated AI leadership
No
Verified conclusion
18Soft2Bet4.8
Named AI product scope
- CRM · AI personalization
Decisioning, prediction and player value
- AI personalization
Yes
Supported conclusion
- AI recommendation engine
Partial
Supported conclusion
Real-time recommendations and next-best actions are available; the proprietary ML boundary remains unresolved.
- Churn / LTV prediction
Yes
Supported conclusion
- AI dynamic bonusing
Yes
Supported conclusion
Risk and player protection
- AI fraud detection
Partial
Verified conclusion
SEON supplies ML risk scoring with operational outcomes measured in Soft2Bet workflows.
- AI responsible gambling
Partial
Supported conclusion
Models surface behavioral change for human review; sensitive cases retain human decision-making.
Conversational, generative and governance scope
- AI chatbot
No
Editorial inference
Customer support uses an in-house AI quality-control system to analyze human interactions; a player- or operator-facing conversational chatbot is not established.
- AI content generation
No
Editorial inference
No generative creation product for promotions, site copy or game content is established. Personalization and recommendation are non-generative capabilities.
- Dedicated AI leadership
No
Editorial inference
A dedicated AI executive or product leader is not established. Applied ML ownership is distributed across product, fraud, analytics and support functions.
Procurement test: model, decision and operating boundary
Start with the decision being automated, then require the exact model owner, data path, production entitlement, target metric, human-control design and change process. The relevant test is whether the contracted system produces a reproducible operational outcome without creating an uncontrolled player, compliance or financial risk.
1. Name the decision
Define the input, output, affected player or workflow, allowed action, prohibited action and accountable human owner.
2. Map ownership
Separate native models, supplier group products, embedded partners, operator models and generic integration capability.
3. Validate on representative data
Set a baseline, evaluation window, coverage, accuracy or uplift target, error cost and segment-level acceptance criteria.
4. Test human control
Exercise review queues, explanations, thresholds, overrides, exclusions, rollback and complete decision-history reconstruction.
5. Test change and drift
Contract model-version notice, revalidation triggers, monitoring, degradation thresholds, incident handling and emergency disablement.
6. Secure data and exit
Define training and inference data rights, subprocessors, residency, retention, deletion, derived artifacts and export on termination.
Adjacent operational records
AI method remains separate from the workflow it serves. CRM, fraud, safer-gambling and reporting records show the broader control surface; they do not turn a rules engine, dashboard or automated workflow into an AI model.
AI and automation FAQ
- Does an AI capability mean that the model is native to the platform?
- No. The capability can be native, partner-delivered, integrated or limited to one product. A named product appears only when the capability is attached to that offer; every provider-level description keeps the remaining ownership and deployment boundary.
- Does Yes prove a production deployment for every operator?
- No. Yes establishes the scoped capability conclusion, not universal entitlement, real-time operation, target-market approval, model ownership or live use in every customer deployment. Contract scope and acceptance testing remain product-specific.
- How should AI model quality be compared?
- Define the decision, outcome and harm first. Require production-like validation with representative data, baselines, time windows, coverage, precision, recall, calibration, false-positive cost, uplift, drift, override rates and segment-level results. A model label or demo is not a performance result.
- Why are fraud and responsible-gambling AI separated?
- They predict and manage different risks. Fraud tooling protects transactions and account integrity; responsible-gambling tooling identifies potential harm and supports regulated interventions. Training labels, false-positive consequences, human review and evidence duties differ.
- Does AI personalization permit fully automated player decisions?
- No. The operator still needs a lawful data basis, eligibility and exclusion rules, market-specific controls, approval boundaries, frequency and value limits, explainability, audit history and a safe rollback path. Sensitive decisions can require human review.
- What should an AI procurement schedule define?
- Name the product and model owner, training and inference data, purposes, prohibited uses, subprocessors, hosting region, retention, entitlement, model version, validation threshold, drift monitoring, human override, incident response, audit access, change notice and exit export.