Business operations automation in the age of Gen AI with Tektonic founder Nic Surpatanu

Gen AI is revolutionizing enterprise automation by offering unprecedented adaptability and intelligence, surpassing traditional frameworks that struggle with complex workflows. To explore these advancements, we sat down with enterprise automation expert and Tektonic’s CEO, Nic Surpatanu, to share his insights on the unique capabilities of gen AI and its impact on the future of business operations.

Henry: Nic, you are a veteran of enterprise automation. What's different now with gen AI compared to traditional automation frameworks like RPA?

Nic: Gen AI is the driving force behind the evolution of Automation, unlocking new capabilities that were not possible before. Previously, software couldn't adapt to user intents or synthesize data as effectively. Gen AI changes that, enabling more complex and nuanced use cases. This AI-first approach can tackle problems that were once managed only by people.

‍Henry: You mentioned the ability of technology to understand intent. Is this improvement purely due to more powerful language models?

‍Nic: Yes, we're moving in that direction. Traditional software required users to format information for machines. With gen AI, the machine handles formatting and translation, reducing cognitive load and making workflows more intuitive.

‍Henry: Tektonic focuses on complex workflows that are hard to automate. How does gen AI and traditional models work together in these scenarios? 

‍Nic: We must look at Gen AI's enablers at different layers. One is the user interface, which becomes much more fluid and intentional, allowing people to navigate it naturally without hardcoded paths. You can switch topics and interact with the AI as easily as a conversation.Another layer is how the system processes data and integrates actions. Gen AI synthesizes information from various sources based on user intent, merging it with system data to extract the most relevant information. Business processes often involve multiple expertise domains and applications that become silos of data, expertise and user interfaces. Working across them is often manual and tedious. The AI can help by simplifying interactions and extracting relevant data, dynamically adjusting based on intent and context.

‍Henry: What about the reliability of automation powered by Gen AI compared to traditional methods? Traditional methods were viewed as brittle. They would often fail due to underlying changes in workflows or data. How does gen AI address this?

‍Nic: Gen AI excels at understanding user intents and mapping them to specific problems. For instance, if I want to create a quote for a client, the AI identifies the relevant entities in the request, like company, product details, and pulls the necessary information from the source systems. However, it struggles with tasks requiring strict business rules and detailed computations, like pricing.The challenge is knowing where to let AI operate independently and where it needs guidance. Gen AI agents need supervision to achieve enterprise-grade reliability, as unsupervised agents can make compounding errors. It's like allowing a new employee to run a complex process unsupervised. Enterprises need clear boundaries and must develop these agents with practice, research, and analysis. This is what we offer to enterprises—a set of tools and templates to help enterprises build and deploy gen AI effectively within these boundaries and for their scenarios.

‍Henry: Given the rise of co-pilots from major tech companies, why should businesses consider Tektonic?

‍Nic: Most co-pilots today are application-specific and leverage data and capabilities within their walled gardens. On the other hand, Tektonic bridges multiple systems, offering a solution capable of handling workflows that cross many applications and systems. Our AI can integrate with co-pilots, enhancing their utility, kind of like agents collaborating with other agents.

‍Henry: You mentioned the complexity of enterprise workflows. Can you give an example of how Tektonic handles these complexities?

‍Nic: Traditional automation often fails when the next step in a process needs additional information or data changes and does not fit precisely the scripted paths. For example, take an automation built to generate service proposals based on customer requests. When services are updated with new descriptions or new services are added, the automation can fail and needs to be fixed. In contrast, Gen AI relies on semantic understanding and mapping between the request and services catalog, which allows the Gen AI agent to continue functioning. This reduces the need for constant manual adjustments and increases the durability of automation.

‍Henry: What are the limitations of gen AI in enterprise automation?

‍Nic: Gen AI is not yet at the level of full autonomy. It shines when working alongside humans that provide oversight to ensure accuracy and prevent mistakes. The AI might misunderstand user intent or misinterpret data, leading to incorrect actions being taken. Given the current state of technical capability, designing complex automation systems with human-in-the-loop processes is essential. The trick is getting the right balance.

‍Henry: While at UiPath, you were specifically in charge of AI-powered work automation. How does Tektonic's approach differ from traditional RPA in terms of AI integration?

‍Nic: Traditional RPA involves prescribed workflows with specific steps, some of which use AI models. Tektonic, however, uses AI to power the workflow itself. The AI works with the user to navigate processes dynamically, utilizing traditional models as tools rather than fixed steps.

‍Henry: An important aspect of the Tektonic platform is that it works with the customers’ own models. What is the "bring your own model" approach, and why is it important for Tektonic?

‍Nic: The "bring your own model" approach allows enterprises to use models from the providers they prefer, ensuring trust and compatibility. Different enterprises have different cloud providers and security requirements. This flexibility ensures that Tektonic can integrate seamlessly with existing infrastructure.

‍Henry: How do you manage the risk of different models performing differently?

‍Nic: We mitigate this risk by specifying tiers of models. We require a minimum tier of capability for specific tasks, ensuring that only suitable models are used. Additionally, we validate the capabilities of any model brought in, ensuring it meets the needs of the solution. While not perfect, this approach ensures satisfactory performance in most cases.

‍Henry: Let’s change gears a bit. Based on your journey with Tektonic thus far, what advice would you give to someone starting a company?

‍Nic: Partner with experienced individuals and organizations. Focus on solving the core business problem while leveraging resources and guidance from those who have been through the journey. Madrona Venture Labs and Madrona have been invaluable in our success.

‍Henry: What's something unexpected that turned out to be valuable in your journey with MVL and Madrona?

‍Nic: Working with MVL has been transformative for Tektonic. From developing the thesis behind Tektonic to validating our product with customers and aiding in team recruitment, MVL has continuously evolved its support based on my needs and the needs of the business. This level of support was crucial to keep the focus on what is most important at this stage of the startup. They also helped me navigate fundraising as a first-time founder. Their network and mentorship have been invaluable in helping us get to this point. And I’m thrilled for our continued partnership ahead.

Founder Spotlight is a series dedicated to bringing you insights into the founder journey, lessons learned from startup leaders, and a behind-the-scenes look at some of the most compelling tech companies in our community.