Fund will invest in early stage, AI-focused startups across all industries and verticals, starting with Rasa

We are investing in Rasa because we believe it is a best-in-class platform for enterprises to develop robust conversational AI

By James Loftus, Alan Du & Rachel X. Zhao

Today, we are announcing the launch of the PayPal Ventures AI Fund, where we are earmarking a portion of our total funds deployed over the next three years for early stage AI investments across all industries and verticals. We are thrilled to launch the AI Fund with our co-lead investment in Rasa’s $30 million series C funding round.

The explosion of artificial intelligence has fundamentally changed the venture capital landscape, particularly with the advent of generative AI. We are starting to see the real application of AI to hard problems as companies build practical solutions that are faster, cheaper and better using AI. We see incredible opportunities in AI across advertising, customer success, risk, compliance, legal, and personalization, and we are excited to launch this vehicle to invest in smart, motivated teams that are disrupting industries across the board with AI.

If you’re an AI founder and you’d like to get in touch with us about funding opportunities, please email: businessplans@paypal.com

---------------------------------------------------------------------------------------------------------------------------------------

Read Rasa’s press release announcing its $30 million Series C funding here.

We are thrilled to be investing in Rasa as our first AI portfolio company. Rasa is creating a new generation of conversational AI to harness the full potential of LLM, and we are backing them because we believe Rasa is the best platform for enterprises to develop robust conversational AI.

Since late 2022, generative AI has captured the world’s attention, spurred largely by ChatGPT’s unprecedented popularity. At the center of it all, advancements in large language models (LLMs) revealed the next frontier of human-machine interface – where linear, mechanical, and visual representation-based manipulations give way to contextualized, intuitive, and verbal communications. LLMs have demonstrated remarkable proficiency in text generation, language comprehension, sentiment analysis, categorization, and even rudimentary logical reasoning. The integration of LLMs into conversational AI will undoubtedly reshape the way we interact with chatbots and deliver more efficient, personalized, and natural conversations.

Many conversational AI tools that enterprises use today run into issues comprehending certain requests because they rely on predefined paths. This type of chatbot must interpret an end user’s utterance for its literal meaning to identify the “intent” before triggering the next action or invoking the appropriate service API. However, when the bot must use inferences and context to identify the end user’s underlying “intent,” they often fail the task. Traditionally, developers have had to rely on machine learning to fix this comprehension gap – but this can be time-consuming. Given recent advancements in zero-shot learning with LLMs, it is now possible for developers to solve the same comprehension problem much more easily.

Rasa has set itself apart as the first and most popular open-source development platform for enterprise-grade conversational AI. Rasa offers a comprehensive toolset that gives developers the freedom to build customized AI chatbots that fit their complex product mix. Enterprise customers must take a measured and prudent approach to incorporating LLMs into their business, requiring conversational AI tools that are robust, flexible, reliable, auditable, and scalable. Rasa delivers an LLM-embedded, multi-model – and even multi-vendor – framework when building their proprietary conversational AI.

Rasa’s recent release of Conversational AI with Language Models (CALM) represents a meaningful leap forward in that it leverages generative techniques to give the chatbots far richer context that allows them to progress the conversation. This also helps eliminate a critical bottleneck in scaling the intent-based approach, where developers traditionally have had to spend countless hours working on intent categorization and edge cases. Perhaps most importantly, unlike a standalone LLM agent, chatbots built with CALM can only send predefined, human-validated responses to the end users. This guarantees zero hallucination. Developers can also choose to allow generative output by incorporating contextual response rephrasing or by integrating enterprise search into the response. CALM strikes a delicate balance between the time-to-value and comprehension of the latest LLMs, and the reliability, transparency, and control of NLU-based chatbots.

Since its founding, Rasa has made it possible for enterprises to build their own, proprietary conversational AI tools bespoke to their unique needs and brand personality. Thanks to Rasa, enterprises no longer have to depend on monolithic “chatbot in a box” or any solution by a single big cloud provider. With the recent release of CALM, Rasa has shown that it is paving the path for enterprises to harness the power of LLMs, while maintaining high standards for business precision and data privacy and security. We at PayPal Ventures are thrilled to support the entire Rasa team on their journey to deliver robust conversational AI to enterprises around the world.

If you’re interested in learning about how Rasa can transform how customers interact with your organization: https://rasa.com/connect-with-rasa

If you’re interested in career opportunities at Rasa: https://rasa.com/careers/