Let’s cut to it. Is DraftKings using AI? Yes—across the stuff you can feel (odds, live markets, personalization) and the stuff you can’t (engineering, service, risk). And no, that’s not just marketing glow. Co-founder Paul Liberman and CEO Jason Robins have both spelled out where AI fits, what’s working, and even which tools are in the mix.
Here’s the thing: when people say “AI in sports betting,” they often picture a robot setting lines in a smoky room. That’s not how it works. At DraftKings, machine learning helps price odds, connect the dots in same-game parlays, and personalize the app’s front door. There’s also a newer layer—generative AI—that’s speeding up code review and some customer support. And if you’re wondering about the “is this safe?” part, there are guardrails and internal policies right alongside the experiments.
Liberman was pretty plain about it: “We’ve been using [AI] for pricing our odds feeds,” he said, and those models “simulate” outcomes and learn correlations for same-game parlays. He also described a long-running push on personalization—from home-page widgets to deposit flows—and even AI-assisted marketing creative and customer service. The team uses Databricks for model management and runs controlled tests before new models see traffic.
Robins, for his part, has said the company is leaning on AI across the shop and that parts of the codebase are now written with AI help. At a Morgan Stanley conference, he noted “personalization through machine learning and AI” and said DraftKings is already seeing a “low double-digit percentage” of code written via AI tools.
And there’s the acquisition trail: DraftKings bought Simplebet (micro-markets, automated in-game pricing) and Mustard Golf (a specialist pricing team), and agreed to acquire Sports IQ (an AI-oddsmaking outfit). Deals like these don’t prove everything, but they do show where strategy is pointed—live, granular, and model-driven.
Odds and live betting: where models meet the money
You can feel AI most clearly when a game is moving fast. Live betting markets need constant price updates; a stale line is basically an invitation for arbitrage. DraftKings’ answer has been to bring key pricing capabilities in-house and feed them with modeling teams. Simplebet’s automated in-play engine, Mustard’s golf pricing models, and Sports IQ’s AI-driven oddsmaking are all pointed at the same North Star: tighter, faster, more available live markets.
Liberman gave the nutshell version: “We’ve been using [machine learning] to understand the correlations between data” for same-game parlays and “pricing our odds feeds.” That helps the book keep up when the ball, puck, or pitch changes the context every few seconds. He also talked through how DraftKings tests models before they “ramp,” with back-testing, split tests, and a model-management setup on Databricks. That’s standard in ad tech and e-commerce; it’s now standard in sports betting too.
To make this concrete: at Databricks’ Data + AI Summit, DraftKings engineers walked through how they surface real-time metrics from the simulation engine using Spark Structured Streaming and Kafka. The point is transparency; traders see how inputs shift the model, then steer pricing with real feedback. Black-box models don’t cut it when a bad correlation can bleed money.
You might ask, “Does this replace humans?” Not really. It’s more like a modern cockpit: the model flies straight and fast, and traders nudge at the right moments. In tennis, baseball, and micro-markets—the most stop-start sports—the cockpit matters most.
Personalization and merchandising: the home screen that knows you
When you open the app, what you see first is not random. Robins told investors that DraftKings has “built out a ton of capability” in merchandising, CRM treatments, and “personalization through machine learning and AI.” The aim is obvious: make the experience feel relevant without making the user do the hunting.
Liberman described the same idea from the product side: personalization touches home-page “widgets and quick links,” suggested deposit amounts, and the flow around them. He added that as the company leaned into personalization and machine learning, customer-acquisition costs dropped “quite a bit.”
Industry coverage of DraftKings’ most recent quarter echoed that product framing. Reports emphasized “hyper-personalized user experiences,” improved merchandising, and churn reduction tied to the company’s data infrastructure and AI capabilities.
Service and code: the quieter places AI shows up
Not every AI win is a headline feature. Some of the best ones save people time.
On engineering, Robins said a “low double-digit percentage” of DraftKings code is now written using AI tools. The guidance inside the company is to treat AI less as a cost cut and more as a productivity boost—ship faster, fix faster, learn faster. You know what? That’s how the biggest compounding effects usually start: small time savings that stack.
On developer workflow, Liberman talked about AI that “identif[ies] problems with our code” as a supplement to code reviews. He also mentioned using AI to generate one-off scripts—say, linking Workday and Snowflake—or to compress batches of media.
Customer service is getting that treatment too. Liberman said they’re using AI to “improv[e] things like customer service”—from routing questions to getting faster responses. That’s less sci-fi than it sounds. Anyone who’s staffed a support queue knows that better triage beats longer hours.
Risk, fraud, and safer play: models with a conscience
Let’s talk responsibility. The company’s job listings describe data-science roles that build models for fraud, anti-money-laundering, and responsible gaming signals. That’s where pattern recognition matters: who’s behaving like a bot farm, a stolen card ring, or simply someone who could use a nudge to slow down? Job specs aren’t press releases, but they do show what a company is actually hiring to build.
DraftKings’ broader responsible-gaming work includes funding state councils and tools like My Budget Builder for setting spending limits and reminders. While those aren’t AI by default, the surrounding initiatives show a push to make safer play more visible and easier to act on. In short: the same data chops that power odds and offers can also power a gentle tap on the shoulder.
The stack behind the scenes
One reason DraftKings can move this quickly is the underlying data stack. Liberman mentioned Databricks for model management; he also referenced scripts that connect Workday and Snowflake—bread-and-butter data plumbing in a modern stack. And at the Data + AI Summit session, DraftKings laid out how Spark streaming, Kafka, and dashboards expose simulation internals to traders in real time. That’s a very modern sports-pricing loop: stream, compute, display, act.
The acquisitions also speak to stack choices. Simplebet owns the machinery for automatically creating and pricing thousands of micro-markets. Mustard’s golf team is a specialist pricing shop. Sports IQ’s pitch is AI-driven odds. Together, they’re ingredients for a book that adjusts quickly and stays up more of the game. If you followed football this fall, you probably felt that uptime.
What AI isn’t doing (yet), according to DraftKings
This part matters because it keeps the hype in check. Liberman said generative AI is used “more internally”—especially for developer efficiency and marketing creative—and that most production code is still written by people, with AI assisting. He also noted that models roll out only after back-tests and split tests, and many run “silently in the background” before they meet customers. That’s the measured, regulated-industry approach you’d expect.
Robins has also pushed back on simplistic takes that AI inherently makes betting more addictive. In multiple interviews, he’s framed AI as a tool for efficiency and better customer experiences, paired with policies across the company. You can disagree about tone—and reasonable people do—but the governance piece is on the record.
Why this matters for the business—and for bettors
For DraftKings, AI looks less like a magic wand and more like a compounding edge:
- Speed. Faster pricing updates keep live markets available longer, which customers notice.
- Relevance. Personalized merchandising and offers reduce the work a user does to find a good bet—or a game they actually like.
- Efficiency. AI-assisted code and support shift effort toward the next feature rather than the next ticket.
On the numbers side, recent coverage tied record revenue and EBITDA to product enhancements, while company commentary emphasized live betting, personalization, and social features as core pillars. Cause and effect is always messy in fast-growing markets, but the direction is clear: better product, better unit economics.
If you’re a bettor, you’ll mostly feel this as smoother live markets, smarter recommendations, and quicker service. And sometimes you’ll feel it as restraint—tools that help you set limits or take a breather. The same math that makes offers relevant can make those nudges timely too.
I like how Robins described the shift: “Everyone is looking at how [AI] can benefit what they do.” There’s a cultural beat there—the move from “I need more headcount” to “I need smarter tools.” It’s small, almost throwaway, but it’s the kind of mindset change that compounds. You can see it in the shipping cadence and the M&A choices.
Liberman said something else worth sitting with: “The AI is only as good as the people behind it.” In sports, that rings especially true. Models handle the grind; people make the calls when emotion and context tilt the field. That’s how you keep edges without losing the plot.
For leaders building with AI: what’s portable here?
A few patterns feel repeatable beyond gambling:
- Tie models to operator dashboards that show their influence in real time. Don’t make humans guess what the model is doing; show it.
- Treat gen-AI as an accelerator for code and service first. Save the moonshots for later. (InnoLead)
- Make personalization a merchandising problem, not just a model problem. The last mile (what gets shown where) is where the magic lands.
And yes—test ruthlessly. Back-test. Split-test. Roll out slowly. Pull back when the data says so. DraftKings laid that playbook out plainly. It works because it’s boring.
If you were looking for a clear “yes” to whether DraftKings uses AI, there it is—grounded in the kinds of details you can only get from people who build the thing and the transcripts that follow them. The short version? Models move odds. Personalization meets you where you are. And the boring parts—code, scripts, service—get faster. That’s how a modern betting app stays ahead without feeling like it’s trying too hard.
