- The chatbot market is expected to generate revenues of $25B by 2030. OpenAI alone expects revenues of $1B by 2024.
- Chatbots can improve CX; 88% of consumers who’ve interacted with them in the past year had a positive experience.
- Chatbots are evolving in three main ways: getting smarter, more conversational and more valuable.
Chatbots are suddenly the most talked-about topic in technology and business, thanks to the November 30 release of ChatGPT. This is an amazingly slick, conversational bot powered by AI. ChatGPT, developed by OpenAI, has so much potential to share rich, well-articulated answers that it has generated an explosion of interest.
OpenAI CEO Sam Altman said ChatGPT achieved one million users in less than a week after its public launch. While the tool is still technically in "research mode," ChatGPT is an example of the power of chatbots to transform numerous business functions.
Chatbots are already widely in use, from Google to online casinos. According to analyst estimates, the market is expected to achieve a compound annual growth rate of 24.2%, with revenues of $25 billion by 2030. OpenAI itself projected $1 billion in revenue by 2024.
Chatbots give businesses an efficient and cost-effective way to respond to always-on customers. And consumers are generally receptive to bots; 88% who had interacted with chatbots over the past year said the experience was positive. Those experiences should only improve as advances in AI make bots more sophisticated and conversational.
Chatbots are evolving rapidly in three major ways: becoming smarter, more human-like and more valuable.
Chatbots often seek to mimic human conversation and increasingly being used to simulate customer service agents during online interactions. Many businesses use them to respond to basic FAQs, such as inquiries about store hours and the status of orders. Chatbots can save customers time by avoiding labyrinthine phone trees just to get a response to routine questions. And they can save businesses money.
Thanks to AI, chatbots are becoming more capable of answering complicated questions – and more personal ones as well. Through machine learning, an AI-powered chatbot can continuously improve service without human intervention. For instance, a chatbot using adaptive AI can improve the accuracy of its replies and learn how to give more personalized responses based on each customer’s needs instead of generic, pre-formulated answers.
One example is AskFrank, a question-answering search engine from Sinch, a Swedish company that provides a cloud communications platform. AskFrank helps Sinch customers find an answer to their questions faster. Each time a customer contacts support, Frank does a quick search on the community page. This reduces the number of times support is contacted and Sinch customers get help faster.
Anthem, a major health insurance carrier with more than 45 million customers, relies on Watson Health (formerly owned by IBM) to answer increasingly complicated patient queries. Its chatbot, Sydney, is reportedly 90% accurate in answering questions about co-payments ("I’m getting a knee replacement. How much does my insurance cover?") and medications ("Does my prescription have any drug-drug interactions?"). Anthem wants to improve Sydney by using AI to go through claims and clinical data to deliver personalized health advice.
This is where ChatGPT comes into play. It’s an example of generative AI, which refers to type of AI that uses unsupervised learning algorithms to create new digital images, video, songs, text or code. Right now ChatGPT is in beta, but OpenAI released it to the public in order to crowdsource feedback and improve its results. People are using it to experiment with a rapidly growing set of use cases, ranging from brainstorming movie plots, to summarizing video transcripts, writing resumes, developing marketing plans, writing code and conceiving of side hustles.
Because the research version lacks access to the internet (it can only access the training data set), ChatGPT cannot provide answers to research questions like a search engine, nor is it ready for businesses to use for CX insights. But once ChatGPT is ready for commercial application (and starts charging for use), it could become the ultimate CX chatbot. Imagine, for example, a retailer answering conversational questions such as "When will Black Panther: Wakanda Forever be available on 4K disc and is it appropriate for children? What kind of reviews has it gotten?" A future version of ChatGPT could potentially scour the retailer’s inventory and data from movie-review aggregator Rotten Tomatoes and provide a perfectly coherent response that addresses all the components of the question.
There's also considerable speculation that ChatGPT could ultimately compete with Google. (It doesn't currently work for local search.) One hypothetical scenario is driving directions. Right now, Google Maps is good for getting you from one destination to the next, in the most efficient way possible. It also offers information along the way, such as whether there's a traffic jam ahead, in which case Google suggests real-time routing alternatives. But a chatbot could take this experience to another level.
A common problem people have driving into Chicago, where I live, is understanding rules about winter tow zones, or figuring out where metered streets end and free streets begin. Wouldn't it be great if I could say to Google Maps, "I want to visit Reckless Records on Belmont. Where is the closest non-metered street, and can you direct me to that street?" A really smart chatbot might be able to tell where you could avoid the meter. But the bot might add something you didn't think to ask: "The closest non-metered street is Ashland, but note that the Overnight Winter Parking Ban is in effect from December to April 1. This means you cannot park on that stretch of Ashland from 3:00 a.m. to 7:00 a.m. or else you will get towed.”
These kinds of possibilities are driving the ChatGPT vs. Google debate, which is now heating up. As Parmy Olson of Boomberg wrote:
I went through my own Google search history over the past month and put 18 of my Google queries into ChatGPT, cataloguing the answers. I then went back and ran the queries through Google once more, to refresh my memory. The end result was, in my judgment, that ChapGPT’s answer was more useful than Google’s in 13 out of the 18 examples.
“Useful” is of course subjective. What do I mean by the term? In this case, answers that were clear and comprehensive. A query about whether condensed milk or evaporated milk was better for pumpkin pie during Thanksgiving sparked a detailed (if slightly verbose) answer from ChatGPT that explained how condensed milk would lead to a sweeter pie. (Naturally, that was superior.) Google mainly provided a list of links to recipes I’d have to click around, with no clear answer.
That underscores ChatGPT’s prime threat to Google down the line. It gives a single, immediate response that requires no further scanning of other websites. In Silicon Valley speak, that is a “frictionless” experience, something of a holy grail when online consumers overwhelmingly favor services that are quick and easy to use.
Still, AI-powered chatbots have a long way to go. They need supervision and human intervention to ensure they don't "spew nonsense" and are free of bias and misinformation. You don’t need to look very far to find instances of AI-powered chatbots coming under fire for going off the rails in one way or another. The New York Times discusses ChatGPT's "misinformation problem" and the dangers of trusting chatbots too much.
As the article points out, chatbots provide answers that sound confident and eloquent, but also reflect the biases of underlying inputs and information when they provide answers. Just because they sound confident doesn’t mean a chatbot is correct. They're also capable of fabricating information by pulling data from multiple sources, a phenomenon known as "hallucination." The more powerful the chatbot, the more potentially powerful the hallucination (here’s a thread about that in the context of ChatGPT).
In a newly published research paper, Deep Mind, a division of Google's parent Alphabet, claims to be addressing the problem of bias and inaccuracy through human intervention. According to Deep Mind, its new AI-powered chatbot named Sparrow uses Google search to source replies to queries. Based on user feedback, Sparrow teaches itself how to provide better answers. Deep Mind relies on a team of people to train Sparrow how to avoid sourcing harmful and otherwise inappropriate content, and how to handle responses to questions deemed inappropriate. For chatbots to get smarter, human training and oversight will continue to be essential.
As chatbots improve they are becoming more human sounding. Their responses sound less like robots and more like real people. Startup Tymely, for example, is developing a chatbot to enable brands to provide email and chat support services in a more empathetic and precise way. And Soul Machines has developed a 3D digital person currently being used on sites such as Calocurb. You can talk or type depending on your preference, and its digital person will smile back at you if you turn your camera on.
Why develop more human-seeming chatbots? Because the more natural and personal the chatbot, the more likely we’ll trust it emotionally, which is different from trusting a chatbot to provide accurate information. Think about how you interact with people. A teacher, coach or boss who provides correct information but in a condescending manner is more difficult to work with because they don't inspire emotional trust – so people might avoid them. Chatbots are the same; tone matters.
At the same time, no matter how human a chatbot seems, people need to know they are talking with a machine or else they may feel "fooled." This is also an ethical issue for businesses using chatbots (and could become a legal one). Here is where human-in-the-loop training applies. Deep Mind is training its Sparrow chatbot to not cross the line into claiming to be a real human. In a recent test, Sparrow answered questions about outer space. But when a user probed further and asked Sparrow if it would go to space, it said it couldn't because it wasn't a person but a computer program. That's a sign it was following the rules correctly.
Delivering More Value
AI can help enterprises by uncovering insights about customers’ needs and sentiment from chatbot conversations. Businesses are under more pressure than ever to mine their first-party data (and data customers willingly share) to personalize services as third-party data become less available. Chatbots can help do that.
For example, Spectrm is a conversational marketing automation platform that brands utilize to build their own chatbots, which can then be deployed on a number of channels. As the bot chats, it learns more about the customer, which triggers a data feedback loop that can help marketers hone their ad targeting strategies and more.
Google is likely to head in this direction by collecting customer feedback data from behind the firewall, what Google calls the unpublished web. Apps such as Business Messages on Google Business Profiles give Google the means to collect this data for brands and businesses. It could then be matched with a business’s location, for example, and provide real-time insights.
Let’s say there's a surge in customer complaints sent via a Google Business Profiles chat about the cleanliness of a particular chain restaurant location, Google could use natural language processing to scan common words in those chats. It could share that data with the brand before those complaints had a chance to spill over into public review spaces. Google could potentially also use the same data as a ranking signal. There are many other scenarios like this.
Chatbots are an intriguing and increasingly valuable way for brands and people to communicate. More importantly, they present an opportunity for brands to deliver better service to more people – by becoming more like humans. While there are certainly concerns to be navigated, the current momentum of AI makes clear that the chatbot future has arrived.