Table of Contents
Here at PolyAI, we’re spearheading a movement to make AI valuable in people’s daily lives. AI can, and should, do more than cut costs. It should solve real problems for real people.
We see a world where customers phone a business and are glad when AI answers the call, because they know that the support they receive will be fast, accurate, and will serve their best interests as a valued customer. We exist to create AI agents that customers find engaging and helpful, not bottlenecks to the service levels they deserve.
Today, I’m delighted to share the next part of the PolyAI story – Agent Studio: The world’s only voice-first omnichannel platform for conversational AI…
How did we get here? A love letter to LLMs, with a catch.
When I started PolyAI back in 2017, I wouldn’t have predicted such an enormous step forward in the field of conversational AI as ChatGPT. It’s not like I was new to the field – my co-founders and I had spent a decade researching dialog systems alongside pioneers in the space like our PhD supervisor, Steve Young. Still, I did not expect the rise of generative AI and the fact that we’d all be using Large Language Models (LLMs) daily – after all, back then, LLMs were mainly just a gimmick for generating Shakespearean-like poetry with little to offer in the way of reasoning or executing tasks.
With ChatGPT, OpenAI moved us into a new era of generative conversational AI. Without getting overly technical about it, generative AI (powered by LLMs) enables AI agents to hold flexible, free-flowing conversations based on its training on vast sets of conversational data, not unlike the environmental data from which we humans learn.
It was a massive breakthrough in the field of AI, and blew the door wide open for those of us working on scalable, lifelike and flexible AI agents that users find engaging and helpful.
LLMs and the enterprise: A love-hate relationship.
As with any big tech breakthrough, it wasn’t long before we started seeing irresponsible and unreliable applications of LLMs. It was suddenly a lot easier for your standard engineer to create an AI agent that could seemingly hold a passable customer service conversation. Eager entrepreneurs jumped on the bandwagon (and who can blame them!), developing ‘thin wrapper’ platforms that put a simple graphical user interface (GUI) over an LLM, enabling almost anybody to develop an AI agent.
The result was mixed. Businesses were now able to develop their own AI agents in-house, but these AI agents were often unreliable and introduced a new risk to brands. Delivery firms were calling themselves ‘useless’ to their own customers and car dealerships were promising to sell brand new vehicles for a dollar.
When you deploy an AI agent with an enterprise customer base, hallucinations or misbehaviors that seemed unlikely during testing become inevitable at scale. With decent guardrails in place, the risk of AI hallucinations decreases significantly – but removing that risk completely requires more than any conversational AI platform offers today. Enter Agent Studio – PolyAI’s platform for reliable enterprise-grade conversational AI.
Welcome to Agent Studio: The world’s only voice-first omnichannel platform for conversational AI.
PolyAI has built and deployed AI agents for many enterprise brands including Marriott, Volkswagen, PG&E, Caesars and MetroBank. Our AI agents speak to millions of customers of these brands, performing transactions like taking reservations, tracking orders, troubleshooting and processing payments.
With Agent Studio, we’re giving enterprises access to our AI agent toolkit so they can build, manage and rapidly optimize customer service experiences in the same way they do for their website and apps.
Agent Studio is packed full of tools to empower enterprises to control and optimize AI agents. Here’s a few of my favorite features…
Voice-first omnichannel
We’ve always taken a voice-first approach to building and designing AI agents. In our experience, focusing on voice first is the best way to create a robust ‘brain’ for an enterprise’s customer service, which can then be leveraged across other channels. In Agent Studio, users can create truly omnichannel AI agents that work off the same ‘brain’ to deliver consistent experience, wherever your customers are.
Complete control over AI models
Agent Studio features an easy-to-use control panel that allows even non-technical users to develop robust rules that force LLMs to say the right thing, and take the right action, at every stage of the conversation. That means enterprises can rest assured that their AI agents are performing reliably and consistently across every customer interaction.
Access to PolyAI’s in-house LLM
Agent Studio users will get access to the world’s first LLM exclusively for customer service – created by our in-house research team. Not only is the LLM fluent in how to speak like a real-life customer service representative (CSR), it’s also fully trained on taking action so your AI agents can perform complex transactions with the finesse of an experienced CSR.
Take control of voice
These days, off-the-shelf speech recognition tools and synthesized voices are pretty good, but they offer little in terms of customization. With Agent Studio, users can fine-tune speech recognition to understand specific brand language and ‘listen out’ for certain types of inputs in each transaction.
Agent Studio also boasts a host of configurable tools for voice output, enabling users to set rules around latency, generate different versions of words and phrases, and cache your favorites for even faster results. You can even give your AI agent multiple voices to bring variation without compromising consistency.
Full transparency on AI-driven decisions
Agent Studio users can review conversations and drill down into the reasoning behind certain utterances or actions. If you need to make a change, you can click directly into your knowledge bases and update speech recognition rules, key information and function calls to improve experience, instantly.
Intuitive interface to optimize CX
With Agent Studio, you can define exactly what your AI agent should and shouldn’t say and do, without compromising on the flexibility of generative AI. Control every aspect of experience from tone-of-voice and pronunciation to conversational flows to API calls.
An unwavering commitment to enterprise voice AI
We know enterprises need control and transparency from the AI tools they leverage. What you see in Agent Studio today is only a fraction of PolyAI’s toolkit for enterprise voice AI. Over the coming months, we’ll be releasing access to more of our platform, enabling our enterprise clients to get more granular control, develop more engaging experiences and join us in our movement to make AI more valuable in people’s daily lives.
Want to learn more about Agent Studio? Learn more and book a demo today.
