Intercom has entered the copilot space, with its new AI assistant for customer support agents promising to enhance accuracy and efficiency.
Dubbed by Intercom as the “only AI assistant that can generate answers using conversation history,” the software provider claims that the Fin AI Copilot helps agents to instantly find their required solutions.
Utilizing retrieval-augmented generation and semantic search, the tool can also provide agents with expert guidance – eliminating the need for manual searches and tab-switching.
Currently available to limited customers in beta but due for general release next month, Intercom is reporting that early users are seeing a 31% increase in support agent efficiency.
The latest Fin solution from Intercom follows the release of the Fin AI Agent last year, a chatbot powered by LLMs and OpenAI’s GPT-4.
The tech firm is currently reporting that the Fin AI Agent has handled over eight million customer queries, saved 670,000 hours of work for support teams, and is resolving 42% of issues on average.
In discussing how the Fin Agent differs from similar chatbots in the sector, Bob Wallis – Senior Product Marketing Manager at Intercom – highlighted the emphasis that the company places on trustworthiness:
Unlike many other AI chatbots, Fin has no problem saying, “I don’t know” and swiftly passing the question to a member of your support team if it can’t find the answer in your support content.
“As a result, you can avoid giving your customers the wrong information, or venturing outside topics related to your business.”
Both solutions are part of what Intercom terms their “AI-first but not AI-only" approach, where the focus is on AI and humans working together to provide the best possible customer experience.
But what exactly separates Intercom's latest AI tool from the rest of the pack? And can it deliver on those efficiency promises?
Differentiators
Varied Content Sources
Fin AI Copilot creates responses for support agents by analyzing conversation history and accessing internal knowledge and external sources such as help centers and public URLs.
Personalized and Controlled