Technology Versus Reality: What AI Actually Changes Inside a Sportsbook
- Kevin Jones
- 2 days ago
- 6 min read
Gareth Crook, SVP of Sports at Pragmatic Play Sports, on the practical use of AI in personalisation, in-play trading, and risk, beyond the industry rhetoric.
AI has become an unavoidable topic in sports betting, but its practical impact remains uneven. Much of the industry conversation still sits at the level of aspiration, while day-to-day sportsbook operations continue to be shaped by latency, risk exposure, legacy workflows, and regulatory constraints. The gap between what AI can do and what can be deployed safely at scale remains wide.
In this Gaming Eminence Q&A, Gareth Crook, SVP of Sports at Pragmatic Play Sports, discusses how AI is being introduced into sportsbook environments incrementally rather than as a wholesale transformation. His focus is on where machine-learning models can support decision-making today, particularly in personalisation, customer profiling, and in-play risk, and where human oversight remains necessary.
The discussion avoids abstract AI promises and instead examines how tools are tested, constrained, and integrated into existing trading and risk processes. Crook also looks beyond AI, pointing to video latency, mobile connectivity, and infrastructure limits as equally important competitive factors that often receive less attention.
Gaming Eminence: From your vantage point at the centre of a global sportsbook operation, where do you see AI making the biggest near-term impact in competitive sports betting?
Gareth Crook: "I would look at this in two parts. First, personalisation has been a core ambition for operators for years; but very few have truly embedded it in a meaningful way. I think AI will accelerate this. From a UI and UX perspective, AI can help surface the right content to a user in real time; and it can do this far quicker and more accurately than traditional rules-based systems.
Through AI models, operators can build a far clearer picture of user behaviour: what sports a customer is likely to bet on; what times of day they place bets; their expected stake range; whether they prefer in-play or pre-match; and which markets they gravitate towards. Those insights can be used to surface the most relevant sports and markets on homepages. They can also be connected to CRM platforms to trigger real-time campaigns through email, push messaging, or SMS.
There is also a strong opportunity on the bet slip and scoreboard side. AI can support smarter upsell; namely, recommended selections, player markets, and bet combinations that align with the sportsbook user’s historical behaviour. For me, this is where AI will make the quickest and most tangible impact in the near term; and it should materially improve the relevance and timeliness of content that users see inside the sportsbook."
Gaming Eminence: Sportsbooks often face a gap between what technology can do and what operational realities allow. How do you navigate that gap when introducing AI-driven tools?
GC: "Yes, I completely agree that there is often a gap between what technology is capable of and what operators can realistically implement within sportsbook operations. We have seen this clearly in the area of personalisation; the ambition is enormous and the potential is there, but the operational realities are far more complex because personalisation has a very long tail and so many different use cases.
For me, this is an iterative process rather than a single transformation. At Pragmatic Play Sports we worked with our internal BI team to create machine learning models that predict the potential profitability of individual sportsbook users. These models use a host of variables with two core objectives: to identify whether a customer is likely to be profitable or not; and to what extent.
Right now, these models do not automate the full risk management process for sportsbook users; however, they provide the platform for our risk team to use the insights to support their decision-making. Our ambition is full automation, but we understand the operational complexities of embedding AI tools into existing processes. So, the journey is iterative: establish the model, allow humans to use it, and then gradually move towards full automation once we are confident the outputs are robust."
Gaming Eminence: Live sports markets can turn on a single moment, a goal, a red card, a weather shift. How can AI and emerging tech be designed to respond to these moments without creating new risks?
GC: "In reality, adoption of automation in sports trading long predates the current interest in AI. In many places, AI has become a buzzword. But when it comes to in-play trading, automation has been embedded for many years. Gone are the days where events were manually traded. Sports are now traded in-play across hundreds of markets that update in real time based on live incident data.
I do think AI tools can make this process even more efficient. One opportunity is around in-play pricing and the aspiration for zero-latency price updates. In an ideal world, a market remains active irrespective of a goal; a red card; a break of serve; or any other key incident. AI models can also be trained to identify the propensity for a sportsbook user to exploit latency – something that still creates operational risk for operators.
Most sportsbooks around the world still impose in-play bet delays and these can be frustrating for users. There are opportunities to remove these completely for specific users. At Pragmatic Play Sports, our platform already supports bespoke in-play bet delays on a user-by-user basis; although, there is no automation or machine learning behind those inputs today. It is simply one example of how the in-play proposition could evolve, and there are multiple others.
Gaming Eminence: Innovation moves fast, but adoption can be slower. What strategies do you use to ensure that AI and other new technologies are embedded effectively across teams and workflows?
GC: "Yes, this is a very good question and a challenge for any business in any sector. I think the answer sits in understanding which technologies should be outsourced to trusted specialists versus which concepts should be built internally with our own development teams. That decision is rarely straightforward and usually involves compromise on both sides.
One example is around scoreboard and match tracker solutions. Innovation has been quite stagnant in this area and only a small number of providers offer genuinely market-leading widgets. While we have ambitions to innovate further in this space, we have chosen to integrate third-party widgets rather than build our own. The main reasons for this are time and complexity. Outsourcing to a provider who specialises in this topic would get us to market far quicker than developing our own solution.
In contrast, on the subject of personalisation, we have decided to develop this internally. A simple starting point is allowing users to favourite their sports rather than rely on the default A-to-Z navigation. There are many other use cases where we believe we can get concepts to market quicker by building the logic ourselves rather than outsourcing.
The strategy is understanding when to choose your battles. Some areas are better outsourced to experts, while others benefit from being developed internally where we can control the roadmap and implement faster."
Gaming Eminence: Looking ahead, which emerging technologies beyond AI have the greatest potential to change how sportsbooks compete and how are you preparing for them now?
GC: There are a host of emerging technologies outside of AI that have the potential to improve the overall sportsbook experience. I would focus on two areas.
The first is sub-second video. New codecs and distribution technologies will enable live video streams with delays far below one second. Sportsbooks globally invest a significant part of their marketing budgets into live video streaming to enhance the in-play experience; yet there is still often a delay between the live stream and the in-play price changes. Prices will usually update ahead of the picture. At Pragmatic Play Sports our live video streaming service is already delivered with a fixed latency of around two seconds from source. Sub-second video will finally allow a sportsbook user to watch the live event in real time with no perceptible delay.
The second area is future connectivity and 6G pre-specs. The majority of sportsbook revenue now comes from mobile, and as a global provider, connectivity by region is a key consideration when assessing load speeds and user experience. Ultra-low latency connectivity has the potential to transform live video streaming quality, support richer front-end data layers, and enable always-on micro-market betting where markets appear and settle dynamically in real time. This is particularly relevant in a high-priority emerging market such as Brazil.