DraftKings used a machine-learning model in 2023 to score customers based on how much money they were expected to lose after receiving promotional bonus bets, according to an investigation published by The New York Times. The automated system examined account activity to target players predicted to incur the highest losses on marketing spend, raising fresh scrutiny over promotional practices in the online gambling industry.
How the DraftKings AI Promo Model Worked
The machine-learning infrastructure evaluated dozens of data points for individual users, according to records and interviews detailed by The New York Times. Data analyst Jayden Butts tested the model in September 2023 across roughly 5,000 online casino players before the company expanded its deployment. The algorithm calculated an elasticity score each week by processing customer play frequency, day-to-day account balances, and loss-to-wager ratios. It also incorporated a separate estimate measuring whether a user was likely to quit gambling altogether.
Former employees told The New York Times that the financial logic identified problem gamblers as the most lucrative targets for bonus offers. DraftKings executives reported to investors that analytics improved promotion-driven sportsbook margins by 13% in 2025, noting that artificial intelligence helped personalize hundreds of millions of dollars in promotional spending.
Contrasting Promotional Targeting with Risk Mitigation Efforts
While the promotional scoring model moved forward, a parallel initiative to identify at-risk gamblers was reportedly sidelined. In mid-2024, data scientist Nestor Hernandez began building a separate risk score model designed to scan money movements, loss-chasing patterns, age, and gender to spot customers heading toward gambling crises days or weeks in advance, according to The New York Times. Six former employees stated that while the data infrastructure existed to run both models, the responsible gaming risk-prediction tool was not completed or deployed.
DraftKings disputed the characterization of its marketing operations. Lori Kalani, the company’s chief responsible gaming officer, told The New York Times that DraftKings monitors customers for risky behavior and declined to deploy certain risk-prediction technology because it lacked sufficient evidence-based backing. The company maintains that promotions target users who exhibit sustained, engaged platform use rather than individuals categorized by expected losses. DraftKings also points to existing safeguards, including collaborations with Mindway AI’s Gamalyze tool and compliance monitoring partnerships with IC360, though former employees questioned whether those measures bridge the gap left by unbuilt predictive safety models.
Keep reading