RIP to RPA: How AI Makes Operations Work

Key Concepts:

• Robotic Process Automation (RPA)
• Intelligent Automation (IA)
• Large Language Models (LLMs)
• AI Agents
• Horizontal AI Enabler
• Vertical Automation Solution
• Unstructured Data
• Data Extraction
• Referral Management
• Computer Use
• Operator
• Return on Investment (ROI)

1. RPA vs. Intelligent Automation (IA)

• RPA Definition: Robotic Process Automation automates manual tasks like data entry and invoice processing. It uses software bots that mimic human clicks.
• RPA Limitations: RPA is deterministic and brittle. It struggles with unstructured data and deviations from pre-defined processes (e.g., misspellings, website changes). It often automates 80% of a task, leaving the remaining 20% for manual intervention.
• IA Definition: Intelligent Automation, powered by AI and LLMs, can process unstructured data, understand context, and determine the best course of action.
• IA Advantages: IA is more reliable and can handle complex, nuanced tasks that RPA cannot. It reduces the need for manual back-office operations.

2. Example: Referral Management in Healthcare

• Traditional Process: Primary physicians fax referrals to specialists. Front desk staff manually input data, check insurance, and review patient history.
• IA Solution (Tenor): Tenor automates the entire referral management process with a user-friendly interface.
• RPA Inapplicability: The complexity and variability of the referral process make it unsuitable for RPA.
• IA Benefits: IA streamlines the process, reduces manual effort, and improves efficiency.

3. Building Intelligent Automation Solutions

• Focus: Start with a specific, repeatable, and manual automation flow within a constrained domain (e.g., industry-specific).
• Integration: Integrate with core systems and understand the context of the industry.
• Example: Automating data entry from phone calls by parsing the call and inputting the information into the system.
• Success Metric: Focus on flows that generate revenue and address constraints on business capacity.

4. Technological Advancements Enabling IA

• LLM Capabilities: LLMs can understand and process unstructured data, enabling more sophisticated automation.
• Browser Agents: Tools like Anthropic's Computer Use and OpenAI's Operator allow agents to intelligently browse the web and take actions.
• Computer Use: A browser agent that understands browser-level actions.
• Operator: OpenAI's upcoming agent with browsing capabilities.
• Impact: These advancements enable startups to leverage AI for industry-specific automation solutions.

5. Two Paths for Building in the IA Space

• Horizontal AI Enabler: Focus on a common component needed across various automation solutions, such as data extraction from unstructured data.
• Vertical Automation Solution: Build end-to-end automation for a specific industry, focusing on a constrained domain and a specific flow.
• Example: A company specializing in extracting key data from unstructured sources, which can be used by other automation solutions.

6. Vertical Automation: Key Considerations

• Target Industries: Identify industries with significant manual work and large back offices.
• Automatable Flows: Determine specific flows that RPA could not handle due to complexity or scale.
• Revenue Generation: Prioritize flows that generate revenue and address business constraints.
• Examples: Automating customer orders by voice, referral management.

7. Market Opportunity and Scale

• Labor Budgets: Many companies have large labor budgets for tasks that can be automated.
• Untapped Potential: Traditional software solutions have not been able to address the long tail of edge cases in these companies.
• IA Impact: Intelligent automation can penetrate markets that were previously inaccessible to software.
• False Comparison: Historical software incumbents do not represent the full potential of the market.

8. Future Evolution (5-10 Years)

• Adoption Curve: Adoption will vary based on the industry's technological sophistication.
• Vertical Automation Advantage: Tailored solutions for specific workflows will drive adoption.
• Focus Shift: Employees can focus on higher-value, customer-facing, or complex tasks.
• Long-Term Opportunity: Early integration and customer relationships will enable expansion into more core tasks.

9. Builder Focus

• Unaddressed Automation: Focus on tasks and industries that RPA could not handle.
• First Flows: Identify the initial automation flows within those industries.
• UI/UX: Develop clean and intuitive user interfaces for these solutions.
• Niche Markets: Explore automation opportunities in niche markets beyond finance and healthcare.

10. Notable Quotes

• "nobody wants to do data entry like nobody wants to sit in the back and read a 100 faxes and try to input that into a system"
• "...if you're able to build an intelligent AI agent specifically for that industry that is tailored to exactly how they do their business it's almost a no-brainer to do it"

11. Synthesis/Conclusion

Intelligent Automation, powered by AI and LLMs, represents a significant advancement over RPA. By focusing on specific, revenue-generating flows within constrained domains, and by leveraging advancements in browser agents and data extraction, builders can create solutions that address previously untapped market opportunities and transform industries with large manual labor components. The key is to identify tasks that RPA couldn't handle and to develop user-friendly interfaces that make adoption a "no-brainer" for businesses.