Introducing Fin: Intercom's Revolutionary AI Chatbot, Based on GPT-4

The arrival of OpenAI's ChatGPT changed everything: the huge implications for customer service were immediately apparent.

We quickly launched a set of AI features for our inbox, applying this technology to drive efficiencies. But the number one question we were asked was "Can ChatGPT just answer my customers' questions?" It was so clear that he understood natural language better than anything we had seen before.

Unfortunately, our early explorations showed that hallucinations were too big of a problem. GPT-3.5, which powered ChatGPT, was just too prone to make things up when it didn't know the answer.

But the recent announcement of GPT-4 changed things: this model is designed to reduce hallucinations. We're excited to share what we've created with GPT-4 in our first tests.

Introducing our new AI chatbot: End

Today we are announcing that we have created an AI-powered customer service bot that takes advantage of this new technology and is tailored to business needs. His name is Fin, and we believe he has the ability to become a valuable partner for your support team.

When we started experimenting with building a GPT-based chatbot, we had several design goals in mind. We wanted to create a bot capable of:

Chat naturally, using GPT technology. Answer questions about your business, using information you control. Reduce hallucinations and inaccurate responses to acceptable levels. Requires minimal setup and configuration.

We think we've succeeded. Our new AI chatbot works "out of the box", instantly reducing your support volume and resolution times.

It can converse much more naturally about support queries than existing bots, bringing the natural conversational ability of modern AI to customer service.

It can understand queries that span multiple rounds of conversation, allowing your customers to ask follow-up questions and seek additional clarification.

Fin engages in natural conversations with customers

We've extended GPT-4 with features and protections specifically designed to provide customer support, where trust and reliability are crucial.

Fin refuses to answer out of scope questions

Fin is designed to only provide answers based on your existing Help Center content, increasing accuracy and reliability. To reinforce this feeling of trust, it is always linked to its source articles, so that your customers can validate the answers.

It's not perfect, as we'll see below, but we believe it's ready for many businesses now.

Fin requires virtually no initial setup time. It ingests information from your existing Intercom or Zendesk help desk, using the power of AI to immediately interpret complex customer questions and apply your help desk knowledge to answer them.

Inevitably, the chatbot will not be able to answer all customer questions. In these situations, he can escalate more difficult questions to human support teams seamlessly. It works effectively within the customer service team, using all Intercom-enabled workflows.

If Fin can't answer a question, she can easily escalate it to a CS rep who can

How we built this AI chatbot

In order to create Fin, we combined the existing technology behind our Resolution Bot with OpenAI's new GPT-4.

We launched Resolution Bot over four years ago, and it works great once set up, but it requires "responses" or "intents" to be manually selected. Unfortunately, we know some customers struggle to overcome this initial hurdle.

"We designed Fin to run using the knowledge that already exists in your Help Center"

For our new bot, we wanted to reduce this friction.

So we designed it to work using the knowledge already created for your help center...