Agentic AI tools begin automating sports and prediction market betting
Software that trades on instruction is moving from racing syndicates into consumer betting apps, according to iGaming Business, raising questions over execution, trust and crowded strategies.
Automated wagering is edging from a niche professional practice towards a mainstream consumer feature, according to iGaming Business, which likens the shift to the arrival of aircraft autopilot more than a century ago.
The publication traces the comparison to June 1914, when Lawrence Sperry flew over the Seine near Paris with his hands raised while a gyroscopic stabiliser kept the plane level, an early demonstration of autopilot technology. Autopilot now runs much of a commercial flight, but a human crew still sets the course and intervenes when required. iGaming Business argues betting is heading in a similar direction.
Under this model, a bettor could hand an AI agent a budget and a set of instructions, such as following a cricket model or an NFL tipster's picks, and let the software research selections, monitor events and place bets within agreed limits.
What agentic AI adds
Automated wagering is not new. Professional racing syndicates already run computer-assisted betting that pairs probability models with software placing bets into tote pools at scale, and Betfair's Exchange API supports automated trading. What agentic AI adds, according to the report, is the ability to work from a broad, plain-language instruction rather than a fixed feed of selections, deciding for itself which data and tools to draw on.
Prediction markets are furthest ahead. Polymarket already supports automated trading, and consumer products are live: Olas launched its Polystrat platform in February, letting users fund an agent and choose or describe a strategy for it to execute on Polymarket, while Forkast's ATLAS lets customers propose and approve trades through Telegram.
iGaming Business expects sportsbooks to come under pressure to offer comparable convenience, and suggests the appeal of hands-off wagering will draw more customers into betting while making specialist tipping and modelling more accessible to casual bettors.
Screenshots are not a strategy
The report is cautious about the marketing surrounding these tools. Profitable trades are easy to publicise on platforms such as X, and providers have an incentive to circulate winning screenshots to sell subscriptions, much as Sperry's stunt was itself a sales demonstration. A single profitable trade, the piece notes, says little about whether a strategy holds up over time. Polymarket's public trading records at least let bettors examine an account's history before copying it.
One practical advantage the report highlights is consistent execution. A recreational bettor might treat similar prices on the same selection as interchangeable, when in fact a shift of a few cents either side of true value can turn an expected loss into an expected profit, or vice versa, once fees are accounted for. Software given a probability estimate and staking rules can recalculate stakes as prices move and refuse bets that fall below an acceptable price, while also enforcing spending limits.
iGaming Business stresses that automation cannot fix a poor strategy and that customers retain responsibility for the underlying approach, much as a pilot still sets the flight path. It cites one contact who uses a Telegram bot to calculate stakes quickly under Australian racing's minimum bet rules, which oblige bookmakers to accept qualifying bets up to a set potential return, but who still approves every bet manually before it is placed.
Professional betting operations, the report adds, will examine execution speed, error handling and how well any tool protects the strategies they have built, since an agent adopted carelessly could dilute an edge developed over years.
The risk of crowding
A further risk is crowding. If numerous AI agents chase the same opportunities, their combined orders can push prices against later entrants, a dynamic iGaming Business compares to Bill Benter's account of horse racing models converging on the same horses and reducing returns for everyone backing them. Copying a successful account's trades is also no guarantee of matching its results, since positions split across multiple accounts can obscure the full picture and a delayed entry price can erase the edge the original trader captured.