How to Build Real-Time Data Streaming for AI | Jacqueline Cheong, Co-founder and CEO of Artie
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In this episode, Jacqueline Cheong, co-founder and CEO of Artie, shares how her company has revolutionized real-time data streaming in an age where less than 5% of streaming projects succeed. (00:02) The conversation explores Artie's journey from a YC startup to closing a $12 million Series A with Standard Capital, focusing on how they solve the complex challenge of moving data between heterogeneous systems in real-time. (04:08) Jacqueline discusses their unique approach to enterprise sales through cold outbound, building automated go-to-market systems without traditional BDRs, and why they've seen customers switch from their own multi-million dollar in-house solutions to Artie's platform. (30:30)
- Main themes: Real-time data infrastructure challenges, AI-driven automation in business operations, enterprise sales strategies for technical products, and the evolution from manual processes to scalable systems
Speakers
Jacqueline Cheong
Jacqueline is the co-founder and CEO of Artie, a real-time data streaming platform. Prior to starting Artie, she worked as a hedge fund analyst at BaliASNI for over three years, investing in enterprise software companies in a long-short fund. She transitioned from the public markets investing world to becoming a software CEO, learning sales from scratch during her YC batch and developing expertise in selling to technical audiences.
Key Takeaways
Focus Relentlessly on Hard Problems
Artie's early strategic decision to focus exclusively on databases as sources, despite lucrative requests for other integrations, enabled them to build exceptional depth and reliability. (26:37) When prospects offered to buy the product if they built Salesforce or Workday integrations, Jacqueline and Robin said no, choosing to perfect their core offering instead. This laser focus allowed them to handle complex edge cases that typically break streaming pipelines, creating their competitive moat. The result was customers saying "what you showed in the demo actually works" - a rare experience in the data tooling space.
Ask Customers for Their Hardest Problems
Rather than showcasing perfect scenarios, Artie tells prospects to "give us your hardest tables and your most problematic data" and actively encourages them to try to break the product. (30:00) This counterintuitive approach builds trust with skeptical technical buyers who have been burned by other tools. By solving the most challenging use cases first, they demonstrate real capability and often exceed customer expectations, turning skeptics into advocates.
Treat Go-to-Market as an Engineering Problem
Artie's marketing and business operations team reports to the CTO rather than CEO, treating GTM automation as a technical challenge requiring engineering thinking. (39:00) They've built automated prospecting pipelines using AI and tools like Clay that replace traditional BDRs, allowing them to scale outbound without human bottlenecks. This approach leverages their technical team's natural optimization mindset to continuously improve and automate sales processes.
Manual First, Then Automate
Jacqueline personally handles new processes manually several times before automating them, whether in engineering, sales, or operations. (38:39) For their first major customer Substack, they used a shared Google Sheet to track hundreds of table syncing statuses, manually updating progress over days. This hands-on approach helps them understand the nuances and edge cases before building automation, ensuring their systems work reliably at scale.
Discovery Calls Are Everything
The most critical skill Jacqueline developed was conducting discovery calls that build trust within the first 5-10 minutes of meeting someone. (66:27) Success comes from asking the right questions to understand pain points, digging deeper into areas customers care about, and providing insights they hadn't considered. She emphasizes never stopping discovery - every interaction should reveal something new about the customer, enabling truly consultative selling to technical audiences.
Statistics & Facts
- Less than 5% of real-time data streaming projects are successful, according to research from Confluent. (17:24) This low success rate explains why companies like Netflix, DoorDash, and Instacart have spent years building these systems in-house with teams of 5-10 engineers.
- Enterprise customers commonly process billions to multi-billion rows of data every month through their transactional systems. (07:32) Jacqueline notes this scale isn't uncommon - every Uber ride, hotel check-in, flight purchase, and app open generates logged data.
- Some of Artie's customers have switched from their own multi-million dollar in-house streaming projects, with teams spending 40% of their time maintaining and running these pipelines instead of building core product features. (23:44)
