AI+Data in the Enterprise: Lessons from Mosaic to Databricks
Feb 26, 2025
auto_awesome
Jonathan Frankle, Chief AI Scientist at Databricks and co-founder of Mosaic ML, shares his insights on transforming AI hype into practical solutions. He discusses the crucial mistake AI startups make when selling to enterprises and how to bridge the gap between research and market needs. Frankle emphasizes the importance of effective storytelling and adapting AI systems to customer requirements. He also highlights opportunities often overlooked by startups and how early wins at Mosaic ML can inform future directions in AI and data intelligence.
AI breakthroughs are increasingly driven by addressing real-world challenges rather than academic research, emphasizing practical applications over hype.
Startups should prioritize effective storytelling tailored to specific industries to engage customers and create valuable case studies for broader traction.
The evolving skill set for AI teams now highlights collaboration and customer engagement, moving beyond solely technical proficiency to solve practical problems.
Deep dives
The Role of Enterprise Customers Versus Startups
Enterprise customers indicate potential business scalability due to their established presence and substantial budgets. They invest in technology for the long term, providing a sense of stability for startups. However, startups can often be the ideal customers for new technologies, as they bypass lengthy procurement processes and offer quick feedback. This quick turnaround allows budding companies to refine their products rapidly based on immediate user insights.
Democratizing AI: The Vision Behind Mosaic ML
The core mission of Mosaic ML was to make machine learning accessible and efficient for all, rather than being constrained by a select few. The goal was to empower users to customize AI technologies based on their unique datasets and needs. As AI technology evolved, especially with the rise of large language models (LLMs), this democratization became increasingly important. The idea is that everyone should have the tools to design AI solutions tailored to their specific business challenges.
Navigating the AI Hype Cycle
Companies are currently at various stages of the AI hype cycle, impacting how they adopt new technologies. Early adopters initially experienced inflated expectations, followed by a disillusionment phase where AI systems fell short of promises. Those at the forefront are now learning to leverage AI successfully by refining their use cases and managing expectations. This iterative learning process is vital as businesses discover which tasks are appropriate for AI integration and which remain challenging.
The Importance of Storytelling in Customer Engagement
Effective storytelling is paramount for startups aiming to capture customers across different industries. While infrastructure products can be horizontal, companies should tailor their narratives to highlight specific benefits for sectors like finance or healthcare. Establishing a strong initial customer relationship can pave the way for additional client acquisition in the same field. Engaging with early customers helps generate compelling case studies and testimonials that resonate with potential clients.
Evolving Skills for AI Teams
The skill set required for AI teams is shifting as the industry continually evolves. While rigorous scientific expertise remains valuable, collaboration, empathy, and customer engagement are increasingly critical. Teams must focus on solving real-world problems rather than purely technical challenges. Emphasizing a culture of continuous improvement and iteration will enable teams to remain relevant and responsive to emerging AI technologies and market needs.
The biggest AI breakthroughs won’t come from Ph.D. labs — they’ll come from people solving real-world problems. So how do AI founders actually turn cutting-edge research into real products and scale them? In this week’s episode of Founded & Funded, Madrona Partner Jon Turow sat down with Jonathan Frankle, Chief AI Scientist at Databricks to talk about the shift from AI hype to real adoption — and what founders need to know.
They dive into:
1) How AI adoption has shifted from hype to real-world production
2) The #1 mistake AI startups make when trying to sell to enterprises
3) Why your AI system shouldn’t care if it’s RAG, fine-tuned, or RLHF — it just needs to work
4) The unexpected secret to getting your first customers 5) The AI opportunity that most startups are overlooking
(00:00) Introduction (01:02) The Vision Behind MosaicML (04:11) Expanding the Mission at Databricks (05:52) The Concept of Data Intelligence (07:42) Navigating the AI Hype Cycle (15:10) Lessons from Early Wins at MosaicML (20:50) Building a Strong AI Team (23:36) The Future of AI and Its Challenges (24:06) Evolving Roles in AI at Databricks (25:55) Bridging Research and Product (28:29) High School Track at NeurIPS (30:39) AI Techniques and Customer Needs (38:22) Rapid Fire Questions and Lessons Learned (42:49) Exciting Trends in AI and Robotics (45:40) AI Policy and Governance
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
