23 Jun 2024 06:16

Advertising & Marketing

How to make a Good Chatbot

Your kid tore his favorite pair of jeans and you need to know if your local store will be open after work so you can pick up a replacement pair. If only you had a personal shopper who could find out what time the store closes.

Instead of you rushing to the store only to find that the jeans are out of stock, your personal shopper would check the inventory ahead of time. If you knew exactly what you wanted, they could ship the jeans to your door so you could skip the trip to the store altogether. And because the personal shopper knew you, they’d naturally ask, “Is there anything else you’re looking for today?”

As long as the personal shopper is a person, that’s a service only a small percentage of customers can get from their local store. But if the interaction is handled by an AI (artificial intelligence) chatbot through a messaging interface, a store can afford to do this for all their customers.


The adoption of chatbot interfaces has been rapid. Apple’s Siri was born less than six years ago, but “she” already has competition from IBM, Google, Amazon, Microsoft, Samsung, Tencent, Facebook and several smaller companies. Surveys show that 60% of consumers have used a voice-activated virtual assistant in the last year, and Google estimates that a quarter of all searches on mobile devices are carried out using voice commands. Even though most bots can’t cope with complex requests, the fact that they work at all creates a great deal of buzz.

In addition to their convenience, conversational interfaces are popular with consumers because we’re wired to think of conversations as interactions with people, not systems or companies. Surveys show that consumers are just as willing to speak to a retailer chatbot as they are to pick up the phone or send an email.

It’s no surprise then that companies are keen to use chatbots for customer service. But turning speech into words, understanding the words, and reacting appropriately is technically very difficult. For most companies, the prospect of building their own natural language processing engine is impossibly daunting. The good news is that there are several conversational AI platforms—which can be used to build chatbots—available. These platforms take care of the nuts and bolts of language processing and let you concentrate on designing the consumer experience.


The most obvious question is: “What are you going to call the bot?” Naming is important. While consumers are comfortable with companies personifying their bots with human names (Siri, Alexa and so on), giving consumers the impression that they’re interacting with a human when they’re not is potentially disastrous. A consumer expecting to talk to a flexible, helpful and understanding person is inevitably going to be disappointed by a bot with limited functionality, even if the interaction isn’t a disaster. Research shows that 73% of consumers say they wouldn’t use a brand’s chatbot a second time if something went seriously wrong in their first conversation.

Despite the risk, the potential of human-like interaction—for example a bot whose “voice” (sassy, measured, serious, youthful, etc.) reflects the personality of the brand—is so great that it’s hard to pass up.

Companies need to figure out how, and to what extent, they’re going to allow their chatbot to use the information they have about individual customers. Ignoring this information isn’t an option—digital shoppers are aware that companies have data about them, and expect them to connect the dots (“I never buy khakis. Why did you send me an offer on khakis?”). A bot that doesn’t “know” the consumer will be unhelpful or obtuse; plus, it’s also vital that the bot uses personal information in a sensitive way. Most people are happy to receive a discount on their birthday, but the sort of data mining that helped one retailer deduce a teen was pregnant and send her offers—that her father saw before she had told him about the pregnancy—proved less welcome. Chatbots need to walk a fine line not often written into your typical one-to-one marketing algorithm.

Along with intimacy and personalization, bots can inject intelligence into the sales process. If your kid’s favorite jeans weren’t available at your local store, a smart bot would check other stores in the chain for stock. If that didn’t result in any hits, the bot would suggest alternative products; based on your value as a customer, the bot might even offer a more upscale option for the same price. And if there were no appropriate alternatives, the bot could offer to notify you when the jeans are in stock again.

Finally, it’s important to recognize that bots aren’t just data consumers. They’re also rich data sources: Consumers are likely to mention birthdays, anniversaries, color and size preferences and more in a conversation. This information needs to find its way back into a customer relationship system to enrich future interactions, rather than being discarded. Bots can even carry out lightweight surveys, gathering customers’ opinions on products or on the service the bot itself provided.


All of these examples show bots hosted outside the store: by smart home devices, in phone apps or on websites. But as with so many other aspects of retail, the digital revolution is blurring the divide between the physical and virtual worlds. Bots may not be as prevalent in brick-and-mortar stores as they are in e-commerce, but they are gaining ground.

They certainly are at Lowe’s. They’ll be deploying machines called LoweBots in 11 San Francisco Bay Area stores this year. Like a chatbot, the Lowe’s robots understand natural language, and can answer questions such as, “Where do I find a cordless drill?” And like chatbots that recognize when a customer is loitering on a website, the robot can spot a person standing bemused at a large display and go over and offer to help.

All of this sounds cool, and improved customer service and engagement are always welcome, but the acid test is how much chat interfaces increase sales and conversion rates. Bots are so new that there isn’t enough data to say. Nevertheless, the huge success of WeChat, with almost 900 million active users in China, and the popularity of chat interfaces among Millennials are trends that cannot be ignored. Companies standing on the sidelines while their competitors experiment with what could be game-changing technology are feeling increasingly nervous. It may not be safe to jump head-first into the water just yet, but it’s certainly time to get your feet wet.


Written by By Ian Dudley, Enterprise Architect at Nielsen

(Visited 1 times, 1 visits today)