Interview with Adam Lewis: A one-to-one conversation with every player, every moment

Turning intelligence into action

The core product, Verso, is designed to bring that vision into an operator’s existing environment without requiring a wholesale technology overhaul.
Lewis describes it as “an intelligence layer that brings real-time understanding, decisions, creation, execution and learning”.
Lewis says Verso is designed to work alongside existing systems with limited integration, allowing operators to begin with a focused use case.
For Lewis, getting the technology right is only part of the challenge. Large operators have competing priorities, multiple stakeholders and extensive project lists. Even when the value of a new solution is understood, getting it implemented requires more than a compelling pitch.
The aim is to demonstrate measurable value against a control group. AxiumAI says a focused Verso deployment can go live within four weeks.
At the same time, operators retain control over how the technology behaves. Verso provides back-office controls covering guardrails, brand tone, messaging parameters and testing frameworks, allowing clients to establish the boundaries before scaling.
Lewis believes that combination of speed, control and measurable results is key to securing buy-in.
According to AxiumAI, a client that launched a high-value use case earlier this year saw a 14% increase in net gaming revenue (NGR), equivalent to approximately €10 per player per month. Across all its products and clients, the company says NGR growth has ranged between 8% and 17% over the past 12 months, depending on the use cases deployed. These figures have not been independently verified.
The commercial argument has become more pressing as operators contend with rising taxation. UK Remote Gaming Duty rose from 21% to 40% on 1 April 2026. A 25% General Betting Duty rate will apply to most remote betting from 1 April 2027; remote bets wholly on UK horseracing will remain taxed at 15%.
“When those changes were announced, they validated our thesis: drive incremental revenue without increasing bonus costs, improving contribution margin and helping offset the impact of higher taxation,” Lewis says.

From catalogue to conversation

Despite the changes Lewis has witnessed during his 26 years in gambling, he believes one aspect of the sportsbook has remained remarkably consistent: the way players navigate it.
The number of markets available on a single match has grown substantially. Where a game might once have offered a handful of options, players can now face hundreds of markets, teams and prices.
For Lewis, “the opportunity is to understand which markets matter to which player, and why, in real time.”
“We’ve become incredibly good at creating choice. But choice on its own doesn’t inspire someone to bet.”
In his view, instead of expecting players to search through markets themselves, operators can use real-time information to surface relevant opportunities and explain why they matter in that moment.
“The opportunity is to move from catalogue to inspiration.”
Verso is designed to create that narrative by combining live feeds from sporting events with each player’s transaction history, which Lewis describes as their “DNA”, alongside news and social media feeds. The system can then generate context around a market as the match unfolds.
A first-goalscorer market, for example, might be accompanied by information that the player has scored in their last two games and takes free-kicks and penalties. The intention is to give the player a reason to consider the market rather than simply placing another option in front of them.
According to AxiumAI’s January 2026 launch announcement, the engine draws on more than 20 billion football statistics and supports over 30 languages. Operators can define the language and tone, from a more informal fan-oriented style to slang used for player names.
Natural language is also central to AxiumAI’s vision, particularly through AxChat, which Lewis says is already live with two European operators.
The proposition is simple: “Ask your sportsbook anything.”
Instead of navigating menus and hundreds of markets, players can ask about a footballer’s chances of scoring, which markets might be interesting, what promotions are available or whether the sportsbook can build an accumulator around a target price.

For Lewis, the important part is not simply providing an answer. It is understanding why the player is asking, what they are trying to achieve and how to provide information that responds to that intent.
His test for every interaction is straightforward: “Is it compelling? Is it inspiring? Does it give the player confidence?”
As he puts it: “Recreational audiences today, they don’t work in menus… they ask questions.”

The player is not the segment

One of the more interesting findings from Verso’s deployment, Lewis says, has been the extent to which player behaviour can change when engagement moves beyond predefined segments. Traditional marketing models are built around historical preferences. If a player regularly backs total goals, for example, an operator may continue to serve them similar markets because that is what their previous activity suggests they enjoy. But a live sporting event is constantly changing, and the opportunities that interest a player can change with it. Someone who typically favours one market may become interested in another as the score, momentum or wider context develops. Lewis says the breadth of behaviour he observed after Verso went live in September 2025 initially surprised him.

“It surprised me at first, but then I thought: why should it?” He believes the surprise reflects the limitations of the traditional approach. By repeatedly marketing to customers according to their established preferences, operators risk learning only what those customers have already demonstrated an interest in. “If we only market to people based on predefined segments, we only learn what they do within those segments,” he says. For operators, Lewis argues, that creates opportunities to broaden engagement and increase betting activity and margin.

The vulnerability question

The ability to deliver more relevant messages at the right moment is also what makes the technology a more complicated proposition.
If information makes a player feel they have discovered an edge, could it be more persuasive than a conventional free bet? And does removing a promotional incentive necessarily make a communication less likely to encourage someone to gamble?
Asked about this, Lewis draws a distinction between offering an inducement and providing information that a fan finds interesting.

“There’s no promotion to it. There’s no enticement,” he says, describing the aim as delivering a compelling narrative that gives players insight and confidence.
It is an important distinction in how AxiumAI positions its product, particularly as operators look for ways to drive growth without increasing promotional spend. However, the absence of a bonus does not, in itself, establish that a message carries no risk. Information can still influence a player’s decision to bet, especially when it is delivered at a particularly relevant moment.
Lewis argues that the same real-time intelligence used to support engagement can also be applied to responsible gambling.

The specific real-time player-risk capability Lewis describes remains in beta, he says. AxiumAI is testing a system intended to manage and protect players in real time. According to Lewis, it is being trained in a semi-live environment, with commercial deployment targeted for 2027. The responsible gambling agents shown on AxiumAI’s website illustrate capabilities in development, the company says.
As Lewis describes it, algorithms would sit within the engagement layer, continuously monitoring behaviour and identifying potential concerns as they emerge. The intended functions include reducing or suppressing communications, changing the nature of an interaction or escalating a player for review, depending on the parameters set by the operator.
“In the middle of the night, human teams are in bed. AI never is,” Lewis says.
The effectiveness of that approach, however, depends on the quality of the signals being monitored, the thresholds applied and the actions operators choose to take.
Lewis says operators set the market-specific compliance parameters within which AxiumAI’s technology operates. In Great Britain, licensed operators remain accountable for relevant outsourced activities; suppliers may also have legal obligations of their own.
Those restrictions can be highly specific. In Germany, the 2021 State Treaty on Gambling prohibits combining live scores from a sporting event with advertising for bets on that event, with an explicit exception for displaying live scores alongside betting offers on an operator’s own website. Lewis says a client there omits calls to action from in-play push notifications, instead providing sporting context and narrative without odds or a direct prompt to bet. The absence of odds or a call to action does not, by itself, establish compliance with all applicable advertising rules.

Independent research provides further context for the debate. An Australian randomised experiment, first published online in Addiction in March 2026, analysed self-reported data from 227 regular sports and race bettors over two weeks. Those who opted out of direct marketing reported 23% fewer bets, 39% less money wagered and 67% fewer short-term gambling harms than the control group. Many participants were excluded after randomisation, mainly because they did not provide proof of opting out, and the analyses were exploratory. A broader report from the same Gambling Research Australia programme identified direct marketing of inducements, such as bonus bets and matched deposits, as carrying the greatest risk of harm.

The findings are relevant to the discussion around promotional marketing, but they do not prove that insight-led messaging is harmless. The study did not separately test communications that provide sporting information without an offer.
The question is therefore not simply whether AI can make engagement more relevant. It is how operators choose to use that capability, and whether player protection can be made as responsive as the technology used to drive commercial engagement.

A wider role for the intelligence layer

While engagement remains AxiumAI’s immediate focus, Lewis sees potential to apply the same underlying technology to other areas of an operator’s business.
One client that began with Verso is exploring how the intelligence layer could support responsible gambling and know-your-customer processes, with payments and compliance among the potential applications over time.
For now, however, Lewis is clear about where the company’s attention lies.
“We want to dominate the niche before we expand.”
The company’s hiring plans reflect that approach. New recruits will predominantly be engineers working alongside clients, supported by a lean core technology team. Lewis believes AI has changed what small, highly skilled teams can deliver, allowing them to move faster than traditional technology structures within large operators.
That also shapes the relationship AxiumAI wants to build with its clients. Rather than simply selling a fixed product, the team can work alongside operators to identify and address the highest-value problems as they emerge. 

Being there for that second

For Lewis, the growing interest in prediction markets reinforces a wider shift towards immediacy. Players increasingly expect experiences that respond to what is happening now, rather than requiring them to navigate a static catalogue.
Sportsbooks, he argues, need to be there “for that second”.
His advice to operators considering AI is to begin with a focused use case, establish clear measures of success and prove its value before expanding further.

As operators look for new ways to grow in an increasingly demanding tax environment, the next stage of sportsbook engagement may be defined by how effectively they can speak to each player individually, and how carefully they decide when they should.

Article written by Elisabet Johansson