Automating customer support agents
MavenAGI is a new software company for the AI era. They recently launched an AI customer service agent, built on the flexibility of GPT-4, which a number of companies like Tripadvisor, Clickup and Rho
Many efforts have been made to use chatbots for customer service, but the experience has by and large been unsatisfying. Bots often misunderstand the question, give unhelpful answers, or sound like robots. AI models were not advanced enough to give human-quality answers.
That changed with GPT‑4. CPO Eugene Mann says, “GPT‑4 was the first time we looked at something that looked like human intelligence and could solve that last-mile problem of communication.”
MavenAGI works by training GPT‑4 to the context of customer service in three stages:
- Maven processes and ingests large volumes of diverse content regardless of format. This includes information from knowledge bases and customer interaction logs from platforms like Salesforce, Zendesk, Freshdesk, Slack, SMS, etc.
- Maven integrates into CRMs and enterprise APIs to understand the context of the user to personalize responses and take actions on behalf of the user. Combining content, context and APIs, Maven leverages GPT‑4’s ability to adapt to each enterprise’s unique tone of voice and uses GPT‑4’s reasoning capabilities to answer questions, take actions and suggest follow ups.
- Maven then self-evaluates, critically analyzing its answers to make sure it’s confident that it’s providing the right answer and action.
“One of the advantages of this generation of AI versus the last generation models is actually you don't really need that much data for any given customer,” CTO Sami Shalabi says. “Evaluation at scale is the magic sauce. The Maven approach has been validated with over 1M customer interactions. Our platform provides the tools that the head of support, CX, and product need to confidently answer any question their customers throw at it.”