How we defined the next generation of Smith.ai’s live-staffed AI Chat

Smith.ai exists to help businesses focus on what they do best. We help them get new customers, keep their existing customers happy, and provide peace of mind. 

To accomplish this, we deliver world-class customer engagement experiences across voice and chat by blending humans and AI.

Today, I am thrilled to announce Smith.ai’s next chapter, powered by generative AI.

Our next-generation chat — powered by advanced large language models (LLMs) and trained on your business’s data — will provide faster, more natural, and more helpful chat experiences to your customers.

Smith.ai still leverages our North America-based agents to supervise the AI software and intervene when a conversation becomes overly complex or requires the human touch. Our AI and humans working together is a core benefit compared to other chat solutions on the market — which require you to have your own support staff.

Why Smith.ai built a new web chat product 

Smith.ai embarked on this new web chat project because recent advances in large language model technology allowed us to deliver on a brand promise we’ve been working toward for years. 

When Smith.ai was founded, the launches of new AI platforms were major steps forward in the technology world. But we felt our desires still weren’t met in how we could have customers successfully interact with AI agents in a way that felt similar to how humans interact with other humans. 

That's because the technology at that time, about five to seven years ago, was still really limited in what it could do and how much it could understand when interacting with a human. The reason: language — as a medium of data — is incredibly complex. The words used, and the intention of those words, have so much variability in them that it was hard for AI to handle reactions properly. Computers require consistency.

At that time, all AI could do was handle very simple interactions with users — similar to an IVR-like experience. That meant AI bots could only handle a specific set of interactions, largely limited to simple Q&A, and entire sequences needed to be pre-built ahead of time for each conversational turn. It only allowed for a very linear path. 

The AI didn’t understand the context of the entire conversation or use that context to support future responses.

Generative AI is changing the game. Advancements in large language models now allow AI to understand and speak natural language much better than before — enabling product developers to deliver conversational experiences that are more human-like. That’s exciting to us because we care about the helpfulness and validity of our chat outputs. 

Next-gen AI web chat

The AI transformation we’ve seen over the past few years (and the massive advancements in generative AI in recent months, like with Open AI’s Chat GPT) have allowed us to work toward a new and more progressive AI chat that works mostly on its own with only necessary agent supervision. It allows us to better deliver on the brand promise we’ve wanted to make since the start of Smith.ai. 

With generative AI, we can deliver more conversational experiences that bring the power of AI and humans together in a way that feels more natural than ever. 

We do this by infusing the AI with just-in-time context from within the chat conversation along with other external content. With access to more data sources — and knowing how to provide natural answers using that data — the AI can function much more autonomously now, allowing us to leverage humans when they are truly needed and not just patch gaps in the AI’s knowledge or understanding. 

The biggest takeaway in all of this is the quality of the AI responses and how AI can successfully fulfill customer needs on its own more of the time. 

Data sources that train the AI

To provide better answers to customer prompts and questions, Smith.ai pulls data from a variety of sources. One important piece of our new AI is the ability to steer a custom large language model with your business’s data. 

Our AI ingests data from the website where the chat resides and can consider that information as part of its responses. That becomes incredibly powerful in providing coherent, accurate, and business- or product-specific responses — and delivered in a natural way.

Right now, the main data source is the business’s website, but soon other sources like CRM data and collateral like sales sheets, product brochures, and technical help docs will inform the AI behind the chat widget, too. With access to more resources, the AI can leverage more content (and context) when determining the right output response. 

Combining free-flow chat conversations and structured workflows

Smith.ai’s updated AI functionality handles free-form communication with structured interfaces like playbooks. They live alongside each other, and we can switch between them seamlessly. 

This means that our AI chat can engage in free-flowing conversation to answer questions and then capture lead information quickly after qualification with task-specific workflows.

Overall, the customer benefits as the AI conforms to them instead of the other way around.

We can deliver way more enjoyable experiences now since the AI is doing a lot of the heavy lifting. We can source information from deeper knowledge bases and deliver more accurate chat replies.

How Smith.ai uses AI and humans together

Conversations are made up of a lot of back-and-forths, some that are knowledge-sharing while some that require coordinated actions; things can get really complex. There's nuance, complexities, and improvisation required. That’s why an AI-only solution can sometimes be a challenge — and one that doesn’t work for many businesses. 

Smith.ai continues to use its live, human agents to play the role where they’ve always excelled. We use AI to handle all the automated parts of conversations, and the agents can intervene and spend time handling more complex situations.

By taking time off their hands from the repetitive and mundane, our agents can enter the conversation later and only when necessary. This could be making sure a lead is actually qualified, validating the AI’s responses, or adding a layer of empathy to a conversation that simply needs the human touch.

Looking ahead

Smith.ai is excited about the next-generation AI used in our 24/7 live-staffed web chat service that continues to offer the best of both worlds. By combining our new AI with our live, human agents, we can handle more chats, have more meaningful conversations, and provide a better experience to end-users. This ultimately benefits our clients in an effort to make sure customer needs are met through each and every chat interaction. 

As Smith’s Head of Product, I’m constantly exploring ways to deliver valuable results for our customers at scale.

Using modern, powerful generative AI technology like large language models allows Smith.ai to deliver that.