Article illustration 1

When Carson Szeder turned five dollars into over a thousand using an AI assistant for an NFL bet, it wasn't luck or deep sports analytics—it was a sign of a burgeoning tech trend. His creation, MonsterGPT, exemplifies the race to fuse artificial intelligence with America's exploding online sports betting market, which saw over $150 billion wagered last year. Szeder's tool employs Retrieval-Augmented Generation (RAG), scraping real-time data from the web to inform its projections, and claims user success rates hitting 56-60% versus standard 52% odds. Priced at $77/month, MonsterBet represents a growing cottage industry aiming to turn AI into a gambler's edge.

The Technical Hurdles: From Tips to Transactions

While AI-powered tip services like FanDuel's Ace chatbot or Rithmm ($30/month) offer analysis, the holy grail remains agents that autonomously place bets. This proves technically fraught. Tom Fleetham's experience with blockchain platform Zilliqa's horse-racing agent Ava underscores the challenge: "She had good analysis, good results. Where it got hard was actually trying to place the bets... It took forever. We gave up."

The friction lies in integration. Most traditional sportsbooks and banking systems aren't built for API-driven, automated betting agents. Consequently, fully automated services pivot towards crypto-native ecosystems:

  • Crypto Wallets as Enablers: AI agents struggle with traditional banking APIs but can operate cryptocurrency wallets, making crypto-accepting sportsbooks and prediction markets like Polymarket primary targets.
  • AgentKit & The Speculative Frontier: Coinbase's AgentKit framework envisions agents handling transactions from flights to crypto trades—and bets. Lincoln Murr, an AI PM at Coinbase, notes early use cases are "speculative in nature," highlighting projects like Sire (formerly DraiftKing).

Complex Models: DAOs, Tokens, and Pivots

Sire operates as a Decentralized Autonomous Organization (DAO), pooling user funds converted into stablecoins. Its AI agents then place bets across decentralized platforms, redistributing winnings minus a "performance fee"—discounted for token holders. CEO Max Sebti pitches it as a "hedge-fund like product." Meanwhile, startups like Memetica (from QStarLabs) showcase the volatility of this nascent space. Initially focused on crypto-token agents for betting, it pivoted within months to building promotional AI agents (e.g., Divinia for Divvy.bet) that hype sportsbooks on social media rather than placing bets.

The Reality Check: Hype vs. Profitability

Despite bold claims, evidence of AI agents consistently beating the market remains elusive:

  • YouTube promoter Siraj Raval's WagerGPT ($199/month) claims automated betting across 40 sportsbooks but users report inactivity and disappointment ("Waste of money," says one).
  • Coinbase's Murr admits: "How profitable these agents truly are, I don’t know."
  • QStarLabs CEO Yang Tang observes enduring old scams repackaged with AI buzz, like tip services sending conflicting picks to different user groups to guarantee some winners and retain paying customers: "All of these tricks are well known."

The convergence of AI agents and sports gambling is a fascinating technical experiment, pushing boundaries in RAG, agentic systems, and crypto integration. However, it also serves as a stark reminder: in high-stakes domains driven by probability and house edges, technological sophistication doesn't automatically equate to sustainable profit. As the infrastructure for agentic transactions matures—potentially led by frameworks like AgentKit—the true test will be whether these systems can move beyond promotional hype and deliver genuine, verifiable alpha in a fiercely competitive arena. For now, the house still holds most of the cards, silicon or not.

Source: WIRED, supported by Tarbell Grants.