Automating Professional Services: Caddi CEO Alejandro Castellano on Using AI to Tackle Admin Work

Peter Ho welcomed Alejandro Castellano to the podcast to discuss how he uses technology to make operations and businesses work smarter. Alejandro shared his background, starting with a childhood dream of being an inventor in Lima, Peru, leading him to study electrical engineering. He pursued entrepreneurial ventures early, starting an EdTech company in college where he learned the basics of running a business. This first company connected him with business people, leading to an opportunity to join a family office, which he described as his "on the streets MBA," where he learned how businesses grow from one to ten million in revenue across various industries. Despite this valuable experience, he always wanted to build something more disruptive in software, promising himself he would pursue a master's in computer science in the US. He eventually attended Cornell Tech, specializing in computer vision and machine learning, with a focus on entrepreneurship.

After completing his master's, Alejandro had the idea to combine his two experiences: the operational knowledge from the family office, particularly in non-tech professional services, with the exploding potential of generative AI to improve these operations. He applied to the Allen Institute for AI incubator, AI2, choosing it because they had invested in AI startups before AI was mainstream, trusting their approach. Moving to Seattle for the incubator, he began exploring his idea. The motivation to start a company stemmed from his preference for working on something he owned or had a stake in, a mission that led him to seek the "best place to build companies in the world".

Observing bottlenecks in admin and back office work in various businesses, Alejandro saw an opportunity to apply generative AI. He initially thought legal would be a good starting point due to AI's text analysis capabilities. Through conversations with attorneys, he identified common manual admin tasks that technical solutions could solve, but which firms lacked the resources to build. He realized this opportunity extended to other professional services like finance and accounting. This led to the core idea of Caddi: an extremely easy-to-use system that could watch a user's screen and replicate their process, like having a consultant ask to be shown a task and then automating it.

Caddi works by having users record their screen interactions while performing a process; Caddi then documents the steps and replicates the process via software, such as API calls. This approach is different from traditional UI-based RPAs, which can break if the user interface changes and might make errors in decision-making, posing a significant liability in sensitive industries like legal and finance. Caddi's hybrid model starts like an RPA but generates actual code, making it extremely reliable.

A key differentiation for Caddi is its focus on professional services users who are non-technical, often highly regulated, cautious about data, and value their time immensely. Caddi was designed to be customizable, allowing each firm to automate processes exactly the way they work today, rather than forcing them to adopt a standard pattern. They position Caddi as a "new AI employee" that users can teach manual processes to, which Caddi then performs reliably, saving time and cost and enabling employees to focus on revenue-generating activities. Caddi also facilitates process standardization across an organization and provides visibility into automated tasks. While they started exploring legal, their current clients are primarily in the financial services industry, as the pattern of needing operational support for expert practitioners is similar across professional services. Their revenue model is subscription-based with different tiers depending on platform usage, offering unlimited seats to encourage widespread adoption within a firm.

Caddi recently raised $5 million in funding, a process Alejandro described as challenging but ultimately quick, taking about three weeks to find lead investors. The AI2 incubator was immensely helpful, providing diverse advisor perspectives (VC, entrepreneur, technical) and, crucially, a network that provided warm introductions to investors, making the fundraising process significantly easier.

 

Key Takeaways

·         Combine your unique background with technological shifts to identify opportunities. Alejandro combined his experience in operatio