OpenAI’s bold move to release chatGPT just 12 weeks ago has changed the way people do work. Last time we checked, they are at around 100m users per month. It’s safe to say it has already become a staple to the day-to-day of many professionals.

swivl has just launched its new GPT-enhanced features, and we're excited to tell you all about them.

We understand that large language models (LLMs) like GPT can be unreliable at times, providing incorrect information. To make sure that these new features are truly useful, we've tested them internally and are slowly rolling them out of Beta to our customer base — already receiving amazing feedback. Our goal is to provide real features that can help our customers, rather than just hype.

We're excited about the potential applications of AI technology in customer service and believe that these new features are just the beginning. By harnessing the power of GPT, we hope to make customer service faster and more efficient, while improving the overall customer experience.

Let’s get to it. Here’s a sneak peek at what we’ve been working on:

‍

Agent Assist

The first feature is called Agent Assist, to help knowledge workers do their job faster while still providing quality service to their customers.

Support teams can utilize AI to generate and edit an automatic response to a customer question within seconds right within the swivl Inbox.

With this feature, support teams can send faster and higher quality messages, responding up to 10x faster to customers with just the press of a button. This means leads and current tenants can expect businesses to respond immediately and contextually to their inquiries, making for a better customer experience.

“When we first launched swivl, I was surprised their AI bot knew more about Self Storage than what I originally imagined. We got to test the new Agent Assist feature and are confident it will save my team a ton of time and enable us to assist even more leads. swivl is an essential addition for any business looking to streamline their communication and improve efficiency," said Nick Newcomb, COO at StorageMax.

Conversation Summarization

The second feature is Conversation Summarization. It pretty much generates a TL;DR (”too long; didn’t read”) for an entire interaction with a customer.

The power of LLMs is great at many things, especially providing context from a large wall of text, such as a chat transcript. This feature allows support teams and managers to view an AI-generated summary of an entire customer conversation within swivl’s Inbox. This new feature can be leveraged to grasp the full context of the interaction without an agent needing to spend time manually summarizing the conversation after it has ended. Furthermore, agents will be able to see a sentiment score in relation to the interaction, making it easier to prioritize follow-ups.

While the summarization feature is not flawless, it still saves a significant amount of time. We are continually refining the feature to improve its accuracy and usefulness.

‍

Will platforms using GPT be able to answer ALL customer queries?

‍

swivl's new GPT-enhanced features leverage the power of large language models to improve customer service, but it's important to note that LLMs don't inherently know the specific questions and answers of each individual business. Furthermore, as with any machine learning model, there is a risk of "hallucinations" - or to put simply, the machine providing incorrect information.

Our platform is model-agnostic: it can make use of models that we have trained and managed ourselves, or models provided by third parties like OpenAI.

General-purpose large language models are trained on a wide variety of public datasets, which often include little to no authoritative information about a business. To function effectively within a business context, natural language experiences must be able to answer questions by referencing curated sets of knowledge.

swivl addresses this with an internal industry-specific knowledge-base and templates. swivl generates answers with content stored in the knowledge-base, which allows businesses to confirm that responses are accurate and grounded in real information - e.g. a certain facility’s information.

“We believe it’s essential that every organization start to understand what AI can do for them. As we continue to push the envelope with Conversational AI in the industry, we remain committed to providing our customers with the most reliable and accurate technology available. We recognize the challenges that arise from using general-purpose large language models, particularly when dealing with the potential inaccuracies that can arise from a lack of authoritative information. swivl is dedicated to developing innovative solutions that ensure our clients can trust the AI-powered responses generated within our platform,” said Mason Levy, Co-Founder and CEO at swivl.

swivl is heavily invested in exploring techniques to provide a system that can combine the conversational understanding of modern generative models with the accuracy that customers need to trust it.

‍

What’s in store for the future of customer service and communication?

This is only the beginning. We are investing time and resources to experimenting with other potential use cases leveraging the technology currently available.

Based on what we’ve learned so far, here’s what might come next:

• As the technology behind LLMs advances, techniques are emerging to reduce the risk of “hallucinations”.
• AI will become a core part of customer service platforms, and will be used to route conversations to the right team, speed up response times, create content, and identify trends in customer conversations.
• While automation will be the future of customer service, there will still be a need for humans to handle more complex questions, solve problems, and provide a human touch.
• This technology is seeing huge upside and will only get more intelligent over time around picking up trends, being able to pull data out of conversations to understand specific talking points to ensure businesses are able to provide the most human experience through their communication stack. It will help us understand what is truly human instead of just having a conversation with our customers.

‍