How Supernormal brings AI chatbots to life

Artificial intelligence has proven to be a helpful workplace tool over the last few months, but this company wants to take it beyond summarizing and note-taking and bring hyper-intelligent chatbots to life — literally. 

Supernormal, an AI-powered meeting platform, has launched an initiative that allows teams to create and customize conversational AI assistants to sit in on meetings and participate as if it were a fellow colleague or manager. From having deep knowledge on meeting topics to being able to run meetings altogether, the goal of this tech is to ensure that teams in any industry are as supported as they can be.  

"We believe in the future where almost everyone has voice agents on calls that feel like you're talking with a human and can do all sorts of incredibly useful things," says Supernormal's founder and CEO, Colin Treseler. "When we look at the the overall landscape of the industry, this is where AI can be immediately impactful and create more human like experiences"

Read more: Hired by AI? How avatars could revolutionize the recruiting process

The company started by developing its own AI companions — one of which is a conversational bot named Norma — that could sit in on their meetings and answer questions, relay feedback, suggest solutions and hold casual conversations with the other participants. Once that was possible, they created a completely customizable option where companies within their network already using their transcription services could build their own AI companion — or use a pre-made model like Norma — based on their needs. 

For example, companies are prompted to select the purpose for their AI agent from a menu that includes options like inbound sales, customer service, event feedback, surveys and employee engagement. The agent then collects the necessary data to be able to add to the subsequent meeting about that topic.  

"These are fully two directional bots that you can have interactions with that will be able to pull out a ton of meeting context," Treseler says. "They're completely multiplayer, too, which means that they will know who's speaking to it and pull information based on who it is. Anyone can give them commands, tell them to stop or to do all sorts of functional things. It's like having ChatGPT directly in your meetings." 

Read more: Why it's up to Gen Z to lead the AI movement — especially at work

So far, the feedback they've received from the companies that have built their own AI agent — which include household names like Salesforce, Pinterest, Walmart, Forbes and Disney — has been overwhelmingly positive. In fact, according to an internal survey Supernormal launched, 70% of them consider it integral to their workflow. 

"You can use the platform for anything a business needs," Treseler says. "We see a huge traction in the HR side, both from talent management and performance management, all the way through to talent acquisition."

Many have begun using it as a screening metric for their interviews, where candidates are prompted to have a 10- to 15-minute long conversation with an agent where it asks talent introductory and follow-up questions based on their resume. Recruiting teams then add the bot to a meeting where it will summarize the best candidates. More uniquely, a recent small clinic in Texas, has used the agents they create as mock patients to train their customer-facing staff. 

However, while these agents are created to simulate real people they aren't intended to replace or reduce headcount at all, an outcome Treseler doesn't believe to be even possible from these AI agents alone. 

Read more: From resume to first day: This AI tool takes over the recruiting process

"We have a pretty core philosophy internally to always deliver time saved to people," he says. "Our AI agents are a tool, not a person. Just like we have computers and calculators and other things that make our jobs easier, this is just an evolution to help people manage their work and get more done." 

As a precaution, Treseler and his team have also built in systems to prevent bias, which has been a problem with other AI chatbots. None of the agents created using Supernormal's platform can answer questions that have been deemed discriminatory and agents have been trained to understand up to 65 different kinds of accents so as to avoid exclusion. Still, he admits that there is a long way to go with these programs, but that the future is bright. 

"We've already seen that collaboration between employees and agents is creating and it's breaking every benchmark," he says. "I think we will see a world in which what we're able to accomplish with these agents and this tech will be truly remarkable."