Tektonic has announced a new $10 million funding round, led by Madrona and Point72 Ventures. The company is integrating generative, neural, and symbolic AI alongside agents into its core messaging. This integration suggests that symbolic AI might be the real workhorse of their solution, with agents serving as a simple rules engine.

Agents Everywhere

A lot of startups shy away from talking about assistants and copilots due to competition from giants like OpenAI and Microsoft. However, if you want to avoid direct competition with these big players, AI agents are a good choice. If your leadership has experience in robotic process automation (RPA), then agents offer even better odds for securing new funding.

Agents and Symbolic AI

Tektonic’s generative AI and neural elements focus on intent identification and developing a fulfillment plan. The term “Rules” is paired with “Symbolic AI” to recall expert systems and appears designed to add an AI sheen over RPA. Many agents are actually rules engines trying to pass as AI, like Rabbit which lacks any semblance of a large action model (LAM).

I don’t believe Tektonic is devoid of AI or simply repackaged RPA. However, claims around AI agents have often been disappointing from both a technology and performance perspective. Agents are more akin to declarative instructions stating objectives rather than how to achieve them.

What is an AI Agent?

An AI agent has several key characteristics: autonomy (the ability to execute tasks independently), adaptability (the capability to navigate new environments without explicit instructions), and decision-making (the ability to choose between alternatives). RPA, on the other hand, lacks these capabilities. It’s a procedural program with imperative instructions that don’t adapt or make decisions.

Does Tektonic Have Agents?

Tektonic describes its approach as using GenAI Agents to augment employees with contextual information and simplified actions through natural interfaces. The company emphasizes the importance of human interaction, feedback, and decision-making in this setup. They highlight that generative AI agents are unreliable when left unsupervised.

Tektonic’s solution involves using GenAI Agents to assist employees by providing contextual information, guidance, and simplified actions. The goal is for humans to interact with these agents, giving feedback and making decisions.

Does RPA Have a Place in the AI Era?

In the AI era, traditional RPA may find a place as a helper tool for generative AI assistants and agents. Function calling can be seen as an RPA bot being called to fulfill tasks. While AI agents are novel because they don’t require detailed upfront design, there will still be instances where quick, procedural functions are useful.

Symbolic AI is AI when implemented correctly (like in a rules engine), but it’s not inherently AI just because of its implementation method. RPA can complement AI by providing the necessary procedures for tasks that don’t require complex decision-making or context understanding.