🔥Upside NBA Video Chat with Ben Levicki (Cavs/NBA), Jesse White (AWS), Garrett Wang (Cerbrec), Andre Antonelli (Machina Sports) On AI Agents in Elite Sports, Best Practices, Case Studies and More

This week we have the honor to interview a group of experts in AI Agents in elite sports to talk about the latest trends, best practices and case studies related to AI Agents in elite sports.

• Ben Levicki, AI solution architect, Cleveland Cavaliers (NBA)

• Andre Antonelli, CEO of Machina Sports., a leading AI Agent platform for sports.

• Garrett Wang, CEO of Cerbrec, a leading AI Agent platform for sports.

• Jesse White, Sr Partner Sales Manager for US Sports & Live Events, Amazon Web Services

Here are the topics that we covered during the video interview:

1. Best Practices in AI Agent adoption

2. AI Agent’s case studies in elite sports.

3. Balancing human and AI expertise.

4. Trust and Ethics

5. AI Agents - Future direction

You can watch the video interview below by clicking on the Youtube link. You can also listen to the audio interview by clicking on the link at the top of the page:

Here are some of the best quotes of our conversation with Ben, Jesse, Andre and Garrett:

---

Q1. Best Practices in Adoption

• Andre Antonelli:
  “Usually it’s always a good idea to start small, ship fast, and iterate fast as well. That way you can always apply the learnings and create a feedback loop from real usage. Even if it’s just a small group of people testing inside the organization, the growth of improvements goes much higher once you have actual real users. You’re not just engineers working on it—you’re getting feedback from coaches, staff, or fans. Once you have real users, you can improve the prompts, the experience, and the agents themselves. The improvements come much faster when it leaves the engineering team and goes to the people actually using it.”

• Jesse White:
  “Working across so many teams, what I’ve seen is that everyone’s at a different stage of their AI journey, but the best place to start is always small and scale gradually. At Amazon we call it ‘working backwards’—you identify a pain point tied to a real business impact. That ensures the first use case you pick actually delivers value and measurable ROI. If you target the wrong problem, it’s hard to get adoption, but if you hit the right use case that truly moves the needle, you get buy-in across the organization.”

• Ben Levicki:
  “With the Cavaliers, we’ve really found success by creating a business-driven roadmap. It’s about prioritization and alignment—getting everyone on the same page with how an AI project will affect their KPIs and the organization as a whole. We use what we call our ‘core four’ questions: does AI enable something that wasn’t possible before, does it solve a real pain point, does it solve the problem better than traditional methods, and does it improve the user experience internally and externally? We weigh each one by impact and use that framework to prioritize projects. That way, adoption isn’t just a technical exercise—it’s organizational alignment that drives lasting impact.”

---

Q2. Case Studies in Elite Sports

• Garrett Wang:
  “We’ve had real success with contract value analysis for pro teams. Imagine negotiating an upper eight-figure contract—it’s high stakes. Our AI agent acts like a playbook for GMs and VPs. Every couple of weeks, they run updated reports on all the players they’re monitoring. The system ingests the latest information, updates valuations, and flags risks. That means when they go into negotiations—whether it’s free agency, restructuring, or a rookie extension—they’re equipped with a constantly refreshed view of player value. The AI helps mitigate risk and improves ROI on contracts that could define a franchise’s future.”

• Andre Antonelli:
  “From a fan engagement side, DAZN used our AI agents for multimodal gamification during the Club World Cup. The system automatically generated quizzes and posts using live statistics, historical data, and real-time news. It worked across more than ten languages, which immediately expanded reach. The ROI was twofold—time savings for studio teams who normally create that content manually, and the ability to cover far more events than they ever could before. The same technology is now being used by sportsbooks to generate content and conversational betting copilots that are always aware of the latest odds and stats. It’s a clear example of how AI agents scale human capacity.”

• Jesse White:
  “With AWS we’ve seen use cases across multiple sports. The NFL’s Next Gen Stats are powered by AI and ML. MLB is far ahead—teams are using AI for pitch analysis, talent ID from scouting videos, and real-time decision-making on the field. NBA teams are using AI for injury prevention, analyzing video of hip and knee flexion to flag potential injury risks before they happen. Swimming Australia uses our AI tools to optimize training and performance—helping athletes peak when it matters most, like at the Olympics. These aren’t hypothetical; they’re measurable improvements teams are already implementing.”

• Ben Levicki:
  “Beyond performance, AI has proven useful in unexpected areas. Our security team was struggling with managing underground parking, so we tried Perplexity’s Labs. Within two days, using no code, they built a fully functioning app to manage it. That was done by a non-technical manager, not IT. It showed us how general-purpose AI can empower people across the organization, saving budget and reducing reliance on external vendors. Sometimes the transformative case studies aren’t in the obvious places like player development—they’re in everyday operations that free up resources and time.”

---

Q3. Balancing Human and AI Expertise

• Jesse White:
  “The way we see it at AWS, AI should augment, not replace human expertise. Some people think AI should replace junior employees—that’s the dumbest thing I’ve ever heard. The reality is younger workers are often the ones leaning in hardest, using AI in their daily workflow. Experienced staff can be resistant, but the future is about balance. Let AI handle the heavy lifting—data crunching, repetitive analysis—so humans can focus on judgment, leadership, and higher-level decision-making. That’s where the real value comes.”

• Andre Antonelli:
  “What AI really does well is narrow the scope and surface the right data at the right time. Coaches, scouts, and doctors are drowning in data—game stats, social chatter, medical records, wearable feeds. AI organizes that chaos. But at the end of the day, especially in high-stakes decisions like health or player selection, a human expert must make the call. AI supports by reducing noise and flagging what might otherwise be missed, but it never replaces human judgment.”

• Ben Levicki:
  “In scouting, every scout has their own style and voice. That makes it hard to standardize reports and compare across the organization. Our AI tools help normalize those inputs and reduce data overload, especially with advanced tracking data like Hawkeye, which records 50 body joints per player at all times. AI structures that mountain of raw data into something decision-makers can actually use. But the scout’s insight still matters—the AI just helps elevate their observations into a standardized framework so leaders can act faster and with more confidence.”

• Garrett Wang:
  “What we’ve seen is that the best results come when AI and humans form a feedback loop. Take contract negotiations: the AI flags blind spots, trends, or inconsistencies, but the management team—who’ve been doing this for decades—can spot when something doesn’t make sense. Their feedback refines the AI. Over time, both sides improve: the AI gets smarter, and the humans get sharper insights. Remove the humans, and the AI’s value drops. Keep them together, and you get far stronger outcomes than either could deliver alone.”

---

Q4. Trust and Ethics

• Andre Antonelli:
  “We designed our platform from the start for explainability. Every output is traceable—you can see which stats, news, or data points were pulled into the context that led to that AI decision. That makes debugging possible when something goes wrong. And because some teams want tighter control, we support on-prem deployments. For sensitive data like customer or medical records, keeping AI inside your own infrastructure builds trust and compliance from day one.”

• Garrett Wang:
  “The first question teams ask us is always about data security. Running on AWS gives us credibility immediately, but we go further—we show where the data came from, how it was transformed, and how conclusions were reached. Raw data alone isn’t useful. What builds trust is transparency around the reasoning: how the AI connected the dots. That lets teams validate or challenge the outputs, and it builds confidence over time.”

• Jesse White:
  “At AWS we say security is job zero—it’s non-negotiable. Customer data stays within their environment and never flows back to model providers. With tools like Bedrock Guardrails, we filter out profanity, hate speech, sensitive data, and hallucinations. That way, AI doesn’t go rogue or make up answers. But beyond the tech, we advise teams to create regular feedback loops. Practitioners should label data and check outputs so the system gets more accurate over time. Transparency about the input, the output, and the process is what builds credibility with players and staff.”

• Ben Levicki:
  “Our philosophy is ‘built by you, for you.’ We start with non-technical workshops—asking stakeholders to explain in plain terms what they want, how they would do it manually, and what success looks like. Then we build guardrails to reduce variability and ensure outputs stay within expectations. One example: if the AI already knows who Evan Mobley is, we can whitelist or anonymize names to ensure bias doesn’t creep in. That kind of transparency—showing not just what the AI is doing but why—helps people trust the outputs and feel ownership of the solution.”

---

Q5. Future Direction

• Ben Levicki:
  “Five to ten years from now, I see fan engagement as the most transformed area. AI agents will unify scattered fan data and personalize every interaction—tickets, communications, even in-venue experiences. Imagine creating your own ticket package with an AI agent that knows your preferences, pulls together the best options, and buys it for you. Imagine a game-day concierge that meets you on the platform you use most and anticipates what you’ll want before you even ask. The ceiling for fan impact is massive, but it depends on whether we can centralize and unify fan data into something usable. Garbage in, garbage out still applies.”

• Jesse White:
  “I like to say sports should feel like airlines. When you fly, from the moment you wake up, your operating system is taken over—your flight details, when to leave, where to go, what gate. Why can’t sports be the same? On game day, your AI concierge could tell you where to park, which gate to enter, which concession stand has your favorite food, even update suggestions in real time based on traffic or game flow. Hyper-personalization will define the fan experience, whether you’re in the venue or watching from home.”

• Andre Antonelli:
  “The real step change will come when agents are always aware of the full sports context—live stats, news, social chatter, fan preferences. That constant ingestion builds a living brain that can power hyper-personalized fan experiences. Once agents are continuously aware and updating in real time, the number of use cases that become possible—across engagement, marketing, and operations—explodes.”

• Garrett Wang:
  “From our perspective, the biggest transformation will happen in business operations. Areas like risk management, recruiting, and player investment are ripe for disruption. AI won’t replace people—it will superpower them. Non-technical staff will become 10x more efficient because AI handles the complexity in the background. That means every manager, every coordinator, every back-office staff member gets augmented with superhuman capabilities, which changes the economics of running a team.”

You may also like: