SalesGPT: How AI Sales Agents are Warming up Cold Calls

Welcome to this week’s Deep-fried Dive with Fry Guy! In these long-form articles, Fry Guy conducts an in-depth analysis of a cutting-edge AI development. Today, our dive is about the groundbreaking open source project, SalesGPT. We hope you enjoy!

Imagine a world where you call customer service and instantly get to talk to a knowledgeable individual who can answer your questions in an attentive and personalized way.

Or … Imagine your company, with limited labor and resources, having the ability to call all of the clients on your list, qualify them, and interact with their concerns and questions, convincing them why they should be your next loyal customer.

This is the power of SalesGPT, which utilizes the intellect and communication ability of artificial intelligence (AI).

WHAT IS SALES-GPT?

The vision behind SalesGPT is that of an autonomous, open source sales agent which is capable of interacting with potential customers through any written channel, including text message, email, WhatApp, WeChat, Messenger, and many more. It can also be paired with crusty.ai, an extension of SalesGPT, to make AI-powered sales calls. This is able to take place through both inbound and outbound communication. As the founder Filip Michalsky explains, “The vision is really to be able to fulfill the tasks of a human sales representative, but as an AI agent.”

Michalsky is currently working an extension of SalesGPT through a new project called crusty.ai. This is an early stage of entry which helps businesses with a few key use cases which were not able to be solved by SalesGPT alone. One of these might be though personal phone conversations with leads, even lower-tier leads which salespeople don’t have the time to talk to personally. Michalsky explains, “This issue can now be solved, as you can literally talk to all your customers at scale, hyper-personalize the offering to them, and then upsell them to a higher tier.” On the backend, SalesGPT is being leveraged as the chat functionality, and Crusty will be able to use a plethora of advanced voice generation tools to make these calls realistic and seamless for individuals and companies. This application, like SalesGPT, will allow for both inbound and outbound calls, giving companies quite a bit of flexibility.

Engaging with a SalesGPT-powered AI agent closely mirrors the experience of interacting with a human agent. This resemblance stems from the AI agent’s capability to discern the specific stage of the sales discussion the prospective customer is engaged in. It can adeptly address concerns, propose remedies, arrange appointments, and execute various user-defined calls to action.

THE MASTERMIND BEHIND THE PROJECT

Filip Michalsky, the founder of SalesGPT, was born and raised in Prague, the capital of the Czech Republic. In 2012, he journeyed to the United States to play squash at Bates College. There, he studied mathematics and chemistry, with a focus in quantum chemistry. After school, he took a consulting job to see how business works before attending Harvard University to attain his master’s degree in data science. It was here that he really began digging more deeply into AI, particularly in the area of AI bias and fairness in machine learning. Following his studies, he worked on recommendation systems at Fidelity Investments for three years before he began dabbling in startups.

Michalsky has been in the AI space since 2015, and he has been fascinated by the tremendous growth of AI since then. He explains how “At that time, it was hardly possible to generate hand-written digits with these generative models … and fast forward eight years and we have flawless pictures being generated, are on the cusp of high-quality video generation, and have extremely good language understanding and translations from text-to-speech and speech-to-text. So there has just been so much cool development.” Michalsky has been inspired to contribute to this upsurge of AI development through SalesGPT, taking generative AI to another level of human-computer interaction.

PEELING BACK THE TECH CURTAIN

SalesGPT leverages the greatest advances in generative AI. The platform consists of a relatively simple stack that ultimately relies on the underlying ability of the large language model (LLM) being used. This includes foundational models such as ChatGPT, Anthropic, Cohere, and many more. SalesGPT allows the user to swap any of these models to use as the “brain” or driving force to power SalesGPT.

An issue with some LLMs generally is that they tend to hallucinate, or make up information that is not factual or dictated by the user. This can be an issue for individual engagement and factual research, which makes it difficult to rely on these LLMs for accurate information and communication. However, SalesGPT utilizes the orchestration ability of LangChain to organize the backend LLM and agent actions, which can help avoid this problem. One of the benefits of this configuration is that the user can constrain the chosen LLM to talk only about their products in a catalog or a company database, eliminating the hallucinatory worries.

SalesGPT allows for a great deal of customization, not only in determining the LLM but also in the ability to create what is called a “system prompt,” powered by LangChain. This allows the user to constrain and customize the LLM to more deliberately and informatively engage with customers and clients through a longer, more comprehensive prompt. Here, the user can inform the AI agent of the agent’s name, purpose for calling, the stage of the conversation, etc., allowing for the customization of the persona of the AI agent as well as give the ability to provide the agent with the necessary backdrop for an effective conversation.

WHO IS USING SALES-GPT?

SalesGPT is an open source project, so the available use cases are up to the imagination of the companies and individuals who decide to use it. Many popular use cases currently include the application of salesbots to WhatsApp and email. However this tool could be used by more than just sales companies. It could also be utilized by manufacturers, human relations firms, consulting firms, and much more. As Michalsky said, “A lot of people are interested in this from all walks of life.”

Regardless of the company or use case, this tool will allow companies to multiply their customer interaction force by 100X or more, allowing for more deliberate and efficient customer communication. It will also allow companies to qualify leads so salespeople and other humans interacting with customers and clients can focus their attention on more pressing tasks and interactions.

WHAT ARE THE RISKS?

Like all great innovations, SalesGPT comes with its own set of risks and concerns, though none of them seem to have much weight when the curtain is peeled back.

One concern some raise is over whether the usage of AI agents will hurt sales opportunities for companies by turning off potential customers who might not want to talk to an AI agent. However there are a few things to consider that help to resolve this concern.

Firstly, communicating with AI is becoming normal for the younger generation. With the growing development and usage of chatbots and AI character/companion interactions, interacting with AI-powered bots is less intimidating or off-putting than it might be for older generations. So even if some are turned off by talking to an AI agent, it seems this issue will resolve itself as this type of communication becomes more normal.

Secondly, it’s worth noting that these chatbots are very realistic. Unlike pre-programmed bots, SalesGPT utilizes personalized responses and human-like interactive features, making for more seamless, natural communication and problem solving.

Lastly, there is a huge ability for the client or customer to save time by utilizing these AI agents. Michalsky prompts, “You can either go to FAQ pages, wait 30 minutes in the phone call center for someone to talk to you, or you can talk to an AI agent right away who can help solve your problem.”

Another concern people might express is over how SalesGPT will impact human jobs. However, Michalsky sees this concern as a natural one with any technological innovation or change. He said, “People didn’t want cars; they wanted faster horses … but this is about overall productivity in society.” It’s also important to note that this technology might be used in ways that will create more upskilled jobs for human workers, shifting the job landscape rather than eliminating it. The application of SalesGPT will also allow human workers in these industries to prioritize their time and effort in more valuable and important tasks that might require more human creativity and relational ability. As Michalsky notes, “I’m not talking about replacing a salesperson, rather about reducing the tedious tasks that come with being a salesperson.”

WHAT DOES THE FUTURE HOLD FOR SALES-GPT?

SalesGPT has made a promising debut, despite being accessible for only a few short months. The project is actively being expanded upon in meaningful ways, including leveraging the platform's strengths to enhance interactions with prospective customers and clients in a more dynamic and engaging manner.

SalesGPT not only has the potential to increase productivity and customer interaction for companies. More, this project has the potential to transform the entire landscape for sales and customer interactions. If this project continues to grow and reaches large-scale adoption, it will surely revolutionize the way we think about and conduct human-computer interactions in business for the foreseeable future.