LivePerson CTO Alan Gilchrest on Humanizing Chatbots and Fighting Bias in AI

Who is Alan Gilchrest?

Conversational AI has no secrets for LivePerson's CTO, as the tech demands constant evolution. Before joining LivePerson, Alan fulfilled several roles at Amazon over the course of 8 years.

Nowadays, it’s hard to find a business or agency website without a chatbot to welcome visitors or customers. It’s a specific artificial intelligence technology that has seen impressive developments in the last decade or so, with several companies now dedicated to further developing conversational AI.

LivePerson is one such company, providing chatbot technology to several types of businesses and needs.

Looking to learn more about conversational AI, the current state of the technology and how it may develop in the near future, we reached out to Alan Gilchrest, LivePerson’s CTO, to elaborate on the matter.

Spotlight: Most online shops and even company websites use chatbots, but to someone who might not be entirely familiar with how they work, could you please tell us what they do and how they evolved over time?

Alan Gilchrest: In the 90’s, web chat started to become a part of our everyday lives, with people messaging one another from their desktop computers. At the beginning, companies didn’t have a presence here. In fact, our founder and CEO Rob LoCascio invented web chat for brands — that’s where LivePerson got its start!

From there, it became clear that to scale up to meet consumer demand, companies would have to use a combination of bots and people to have conversations with everyone reaching out to them via chat.

As the technology progressed and as people began to text and message on mobile devices all day, every day, customer service and shopping chatbots have evolved so that they can handle asynchronous (not tied to one session) conversations, as well as feel much more natural and helpful than they did in the early days.

Today, you might […] work with a bot to customize a burrito order, book a flight, upgrade your phone or find the perfect Christmas gift. It’s all about having a two-way conversation that gets you what you need.

What about conversational artificial intelligence? What exactly is it and where does it fit within the chatbot infrastructure?

Conversational AI is the technology that lets bots communicate with humans through natural-feeling conversations. Whether through voice-based or text-based interaction, it recognizes your intent, then takes action to resolve that intent. Old-school chatbots were rules-based programs that provided appropriate responses for very particular scenarios.

They would be triggered by defined keywords rather than true conversations. In contrast, Conversational AI combines natural language understanding (NLU), natural language processing (NLP) and machine-learning models to emulate human cognition and engagement.

What makes LivePerson’s Conversational AI “anything but artificial?” Can you tell us more about the advanced technology behind it?

Unlike more generalist AIs that aren’t specialized to handle any one thing, we are laser focused on AI for customer engagement. Our AI is powered by over 20 years of conversation data we’ve accumulated helping thousands of clients, including the world’s largest brands [that] message with consumers at scale. In fact, we now power about a billion conversational interactions a month, which gives us a unique dataset for creating AI.

All of this runs through our Conversational Cloud platform, which has tools for easily building and optimizing bots, analyzing customer intent and sentiment, giving human agents AI-powered recommendations for how to best serve customers and much more. From the data underpinning it to our tools [in which it’s based on] to the outcomes it drives for our customers, we’re creating world-class AI. Fun fact: Fast Company named us the #1 Most Innovative AI Company in the world this year!

Different businesses have different needs and concerns. How do you ensure the quality of the conversations with each client and how do you measure its success?

When it comes to customer engagement, we’ve found that working against a simple framework can help create great experiences. Based on our deep well of data and years of experience measuring and improving conversations, we came up with the LivePerson 4E Framework. This methodology combines 22 different metrics into a set of four easily understandable ideas.

On the consumer side:
Effort: how much effort did it take to get what I need?
Emotion: how do I feel about this interaction?

And on the brand side:
Efficiency: how efficiently did we handle this?
Effectiveness: how well did we handle this?

These four E’s are represented in easy-to-understand formats within our platform and they are paired with recommendations on how to improve against each to make conversations better. Using this framework, you can tell at a glance how conversations are performing and how to improve them going forward.

Speaking of clients, can you please share some of the biggest brands partnering with LivePerson and what were the challenges and solutions for dealing with big data and analytics for enterprises?

We work with hundreds of the world’s top brands across every industry you can think of — Virgin Media, HSBC and Delta Airlines all come to mind. As you can imagine, very different brands need very different solutions. Some hire their own teams of conversational designers and bot managers so they can handle building, managing and optimizing Conversational AI in-house. Others rely on us to run essentially their entire conversational program, from providing agent talent to designing and tweaking their automations.

Our job is to partner with them to understand their overarching business objectives, what their definition of success is and then work with them along their journey to get there. We help our brands develop a truly deep understanding of what their consumers need. Behind every call or text is a human and we help these huge brands make each experience with their millions of customers feel uniquely human and very personal.

One of the challenges behind making things feel personal is that the old ways of creating personalization, like cookies, aren’t really trusted. Consumers clearly want personalization: according to some research we did earlier this year, 79% are more likely to make a purchase from a brand with which they feel a personal connection. But they’re wary of the old ways — 76% say they are wary of brands using cookies. The good news is that, in a conversational experience, you can just ask them questions!

In your experience, which industries do you think benefit the most from chatbot technology and why?

I can honestly say that there’s no one industry that benefits most from Conversational AI and automation. Anywhere you need a conversation to take place, AI can help. To date, we’ve helped retail, travel and hospitality, banking and financial services, and even crypto companies use Conversational AI to scale personalized conversations, just to name a few industries. When you get down to it, conversations are always at the heart of doing business.

While there’s been some great advances in the area like advanced speech recognition, chatbots are still far from perfect. What have been the biggest roadblocks developing this technology and where do you think we’ll see the biggest advancements in chatbots during the next decade?

We still need to do a lot of work to make conversations with technology feel more human, to make conversational experiences that aren't just about automation and solving one-time problems. We want to create brand new ways for humans to communicate with machines in a way that builds true long-term relationships.

How can we make people want to talk to machines? How can we inspire people to think differently about communication, how to be open and learn more about who they're talking to? In the future, AI won’t just talk to humans, it will actually help them think better and do better.

Some of the challenges here are around how AI self-heals and self-learns. We’re working on creating loops where the AI learns and improves on its own based on conversation quality. We’ve made a lot of progress here, like with our Meaningful Automated Conversation Score (MACS), which is a first-of-its-kind way to measure how conversations with bots are going from the customer’s point of view.

This is the kind of signal an AI needs to get better at understanding consumers, reset conversations to a known good state and delegate to other capable bots and humans […] I think self-healing and self-learning are where we are going to see big leaps forward.

Will there ever come a time where people won’t be able to distinguish if they are talking with a chatbot or an actual person?

This is a really interesting question. The issue isn’t when this will happen (it’s imminent), but how we choose to deal with it. At the end of the day, I think we and the partners we work with want to build trust with the people we’re talking with, not trick them. We need to use AI technology to build a better world, not foster distrust or perpetuate the inequalities that are already built into our societies.

To that end, I’d encourage anyone working with AI to take the EqualAI pledge, which identifies actions you can take to fight bias in AI. LivePerson is a founding member of EqualAI, a non-profit that also offers certifications like the EqualAI Badge Program to help all kinds of organizations make tangible steps toward responsible AI governance. You can learn more about that at equalai.org.

Thank you Alan Gilchrest and LivePerson!

Keep up with LivePerson on Instagram @livepersoninc and read about other successful businesses here.