The problem with generalist chatbots

Chatbots, all too often, are conceived of as concierges. And more often than not, these generalist chatbots are not enormously helpful on a website.

The ‘chatbot as concierge’ model fails to recognise that your visitors are typically looking for a simpler, clearer support offering. They need help with their specific intent more than they need open-ended chit-chat. When it comes to your website, then, a valet is more useful than a concierge.

Here’s the problem with generalist chatbots.


What is a generalist chatbot?

Generalist chatbots are those that try to do everything. They’re all-purpose chatbots, acting as a concierge for your website. They try to do it all.

Take Siri or Alexa, for instance. Both are examples of voice-powered generalist chatbots. They tell jokes, perform searches, play music. You can ask them questions about general trivia, the news, even your calendar.

In the context of your business website, a generalist chatbot would go beyond answering queries. It’d greet visitors, help customers with their cart, suggest items, act as a navigational assistant, and more. Provided, that is, it works well, and understands what the user wants.


The alternative: specialist chatbots

Another way to understand generalist chatbots is by looking at their counterpart. Specialist chatbots focus on one thing or area, and they do that area really well. Think a valet, not a concierge. They have a clear goal and they work with your visitors to achieve that goal.

The most common example of a specialist chatbot is an FAQ bot. It doesn’t power chit-chat, and it doesn’t answer random questions or offer assistance throughout the user’s website journey. Rather, it’s trained on a company’s top FAQ and can reel off answers to them quickly and accurately.

Other examples of specialist chatbots vary by industry. For example, a specialist bot in the hospitality industry might offer a flow-based booking service. So, it would offer a limited scope of options for the user to select and then make a booking accordingly. It wouldn’t invite open conversation or veer from the script.

In general, specialist chatbots are narrow knowledge or utility windows with a set focus. While generalist chatbots skim every task, specialist chatbots are honed in on their one primary function.


The struggles of generalist chatbots

Generalist chatbots, in theory, are great. In practice, however, there are three main problems.

  • 1. They’re complex

It’s difficult to make a generalist chatbot that works. And harder still to make one that works well.

  • 2. They’re less helpful

Compared to their specialist counterpart, generalist chatbots go into less detail. As such, they offer less in-depth support.

  • 3. They lack clarity

Generalist chatbots, supposedly, can do anything. But it’s not clear what the limitations are. This either sets high expectations or fails to help customers use them. It’s easier to use a chatbot if you know what it can do.


Generalist chat-nots

When it comes to chatbots — of any scope or type — it’s imperative that they can reply to users effectively. When they don’t, they’re chat-nots. They shut down conversations, rather than engage in them.

Little is more frustrating than repeatedly receiving ‘try again’ messages or getting stuck in a loop with a confused chatbot. Particularly when that chatbot is supposed to do everything.

Generalist chatbots are exceedingly complex. Because they try to help with everything, there are thousands — even millions — of possible conversation paths. As such, creating a generalist chatbot that works is incredibly difficult. Doing so takes masses of data and training. Trying to do everything means there are more opportunities to fail.

Specialist chatbots, on the other hand, direct all development attention onto one task. The attention to detail is easier to achieve, meaning there’s less chance of conversation error. It’s easier to understand user intent when it’s focused on one core area of your service. This is because there are only a few things that it could be.


Less helpful

Even working generalist chatbots bring problems, though. In many cases, a generalist chatbot ends up being less helpful for users.

Your online customers visit your site for a reason. They’re there to get a job done. Maybe they want to solve a problem they’re experiencing with your products. Maybe they have a question about your services. Or, they might come in search of a specific item. The list goes on.

They aren’t there for a chatbot that does everything sort of well. They want one that can effectively support them through their specific issue. Often, generalist chatbots are too vague for this. They can offer all-purpose advice. They can help with problems to an extent. But they don’t go into detailed support the same way that a specialised chatbot could.  

When it comes to support and problem-solving, having a specialised chatbot means more detail — and so more help.


Clarity

The problem with generalist chatbots is that they lack clarity from the start of the interaction. When you look at a valet, or speak to one, you know exactly what you can achieve. When it comes to a concierge, it’s less clear.

Generalist chatbots offer non-descript help. It’s not clear how they can help users, or what a user can achieve by talking to one. As a result, your website visitors must take a chance and ask the bot before they know whether it can help them.

“Hi, can I help you today?” Is likely to have your visitors thinking ‘I don’t know, can you?’ or get an answer along the lines of “just browsing”.

Compare this to a specialist chatbot. Instead of offering non-descript help, a specialist chatbot can open with a directed question.

  • “Is there a question about your order I can help with?”
  • “Would you like to book a demo?”
  • “Do you have a question about pricing?”

These questions speak to the ability of the chatbot. They direct the user toward a conversation the bot can handle. They make it easy for users to engage.


Choose a specialist chatbot

Generalist chatbots are cool, but also complicated, elusive and unneeded. They’re nice to have, and arguably, the pinnacle of what we want a chatbot to be. But without artificial general intelligence, they can all to often fall short.

Plus, when it comes down to it, a generalist chatbot is often excessive. It’s better to have a chatbot that answers specific customer needs well, instead of one that can do everything adequately.


Useful links

10 practical chatbot uses in ecommerce

The pros and cons of a flow-based chatbot

What is an intent-based chatbot?