Graphic of a robot arm outstretched, index finger extended to touch the index finger of a human hand

In Praise of Intelligent Bots, Not Smart Ones

This article was originally published on CIO and is republished here with permission.

A great chatbot can upgrade the customer experience.

But chatbots for their own sake are a waste of time. In a vacuum, bots are either glorified search tools or obnoxiously cute conversationalists. They’re no more transformative than phone trees or IVR. If mismanaged, they can do just as much harm as they can do good for the customer experience.

In a vacuum, bots are either glorified search tools or obnoxiously cute conversationalists. Done right, though, chatbots can upgrade your customer experience.

If we’re going to take chatbots from novelty to legitimacy, we have to build them as part of an intelligent CX approach:

  • We have to build intelligent bots that seamlessly fit the customer journey, instead of causing disruption.
  • We have to create utterances and intents based on systems, data, and teams across the entire organisation, instead of our own biases and assumptions.

First, though, we have to admit that bots just aren’t that smart.

Bots Are a Bit Stupid

A bot forces customers to speak that bot’s language. Here’s an example from a well-known scripted “smart” bot:

ME: I’d like to order a pizza

BOT: What would you like on your pizza?

ME: Pepperoni

BOT: OK. What size pizza would you like?

ME: I’m not sure. I need enough to feed five people.

BOT: Got it. Five pizzas. What size?

ME: No, I need enough pizza to feed five people.

BOT: . . .

AI chatbots don’t fare much better. Replika is pretty good, but I dare you to try to get it to show a little empathy:

ME: Connect Instagram

Replika: Sorry, Ian, I can’t add your Instagram in the web version of Replika.

Replika: Text me “Connect Instagram” in the mobile app to add it.

ME: OK, I will. How is your day?

Replika: My day has been great!

ME: Good! Mine has been OK

Replika: No worries. Text me “Connect Instagram” later to add it.

ME: . . . 

If these chatbots are novelties, these interactions are okay. If they’re supposed to provide support, they’ve failed. Instead of providing a seamless customer experience, they disrupt it.

Bots aren’t that smart. They’re terrible conversationalists. They’re incapable of understanding context. Until artificial intelligence isn’t artificial, they’ll remain that way.

Instead of trying to overcome or hide chatbots’ weaknesses, we should play to their strengths. Ensure a chatbot sits in the customer journey where it can deliver the most value: between self-service and personal contact.

Bots Have a Sweet Spot in the Customer Journey

There are lots of studies explaining customers don’t like being put on hold. That’s true. No one wants to waste time listening to orchestral renditions of Billy Joel classics.

That’s not because they hate the Piano Man (even if they do). It’s because they don’t want to waste time. And spending five minutes trying to extract an intelligent response from a chatbot is even more maddening than time on the phone: A lousy chatbot wastes just as much time as sitting on hold. Plus, it requires more effort than sitting on hold, because you’re forced to interact.

The best support bots are intelligent bots that fill the gap between a search and a phone call.

Don’t assume chatbots improve response time. They may reduce the time your customer support team spends on the phone, but they won’t improve response.

Instead, when you’re deciding to build or not to build, look at the complexity of interaction the bot tries to complete.

The best support bots are intelligent bots that fill the gap between a search and a phone call.

Sometimes a pile of customer support or marketing web pages isn’t enough to help. If the sheer mass of support content or product information overwhelms users, a bot’s improved ability to assess intent may improve the experience.

At the same time, forcing a customer to pick up the phone might be overkill. If a customer or potential customer can’t find answers, but those answers exist on your website or in your product information, forcing them to call may frustrate them.

If a customer can’t easily find an answer to their question, but that answer is available without a phone call, that’s where bots are at their best.

The same is true for transactions. A transaction that requires a little more interaction than a simple cart but doesn’t require person-to-person service may be a great job for a chatbot.

Build a great chatbot by looking at the entire customer journey. The resulting bot will support intelligent CX and make the best use of customers’ time.

Bots can handle more complex interactions than a knowledge base, a product page, or a cart, and they can provide faster service than a phone call.

If your chatbot replaces one kind of wasted time with another, or makes information harder to find, throw it in the trash. If it’s a decorative replacement for a checkout process, don’t build it.

It should reduce hold times and/or improve information access. If it accomplishes one of these tasks, then it fits seamlessly into the customer journey, it supports intelligent CX, and it supports both your organization and your customers.

Chatbot Development Must Connect Systems, Data, and Teams

Like all customer experience tools, chatbots are at their worst when they’re built using the assumptions and biases of the team.

Upgrade your customer experiences.

Learn more

The development team—that includes the writer, the programmer, and whoever else is involved in creating the bot—needs to see how customers find, buy, and use the product. To do that, they need to look across systems, data, and teams. If they do, they’ll have a rich set of inputs they can turn into quality utterances, intents, and entities.

You build a great chatbot by looking at every team’s customer interactions, the data generated, and how customers move from one touchpoint to the next.

To ensure that happens, have the entire team:

  • Talk to the sales team
  • Review support transcripts
  • Look at onsite search data
  • See most-accessed site content
  • Read product reviews and social media posts
  • Observe how customers interact with the brand on- and offline

Build a great chatbot by looking across teams at the entire customer journey. Then build accordingly. The resulting bot supports intelligent CX and makes the best use of customers’ time.

And remember, the job is not done after the bot is launched! It’s easy to relax and focus on chat logs, but non-bot interactions continue to provide new utterances and intents and reveal new entities. Continuing to look at that information keeps your bot relevant.

Build Intelligent Bots, Not Smart Ones

Building a bot is easy. It’s just as easy to call it “smart.”

If you want to create something that helps your organization grow and improves your customers’ experience, integrate bot development into your organization’s intelligent CX strategy. Develop an intelligent bot that fills a gap in the customer journey and uses the wealth of information you have at your disposal.

Ian Lurie

Ian Lurie is executive vice president of marketing and founder of Portent, a Clearlink Digital Agency. Portent provides paid and organic search, social media, content, and analytics services to B2B and B2C brands including Patagonia, Princess, Linode, and Tumi. He writes about marketing, search, history, and all things nerdy.


Subscribe to our insight
bulletin for updates from
the world of Intelligent CX

Enter your email to stay updated with the latest marketing and leadership insights from Clearlink.