Skip to content
Artificial intelligence

Can we chat? Chatbots are hot

Bill Keenan  Senior Director, Research & Development

Bill Keenan  Senior Director, Research & Development

It’s a bit like déjà vu, but chat technology is once again hot.

Whether it’s within Snapchat, WhatsApp, Skype, Slack, Facebook, Allo, or your local chat tool, people are spending lots of time messaging with each other, and the tech world has taken notice. Why? Because that’s where the people are.

At the same time, app developers are finding it increasingly hard to get people’s attention, as users tend to limit the number of apps they employ. So if you’re a company offering services via an app, forcing the user to switch to that app can create just enough inconvenience that you lose opportunities. Plus, the immense number of apps available makes it hard for users to discover a new app.

So if people spend a lot of time in chat/messaging apps, what’s a service provider to do? Go to where the people are and use bot technology to interact with the user and offer services there!

Multiple use cases for chatbots

The use cases for chatbots are diverse:

  • chatbots-graphicCustomer service: a chatbot can be the first point of contact in a web-based helpdesk chat session. It can greet the user, get pertinent information and answer a range of frequently asked questions. If it can’t provide the answer, the bot can intelligently route the call to an agent. (See examples at right.)
  • Search and question answering: Ask Siri , your intelligent personal assistant and knowledge navigator; she knows!
  • Information delivery and alerting: Why should a user have to leave a messaging app to get important information. A bot can consume alerts or news from other services and push it to the user in messages.
  • Internet of Things Bot: The user can interact with their smart homes and devices.
  • Shopping: Bots are being used to sell fashion, shoes, flowers, and more, and they can provide guided recommendations.
  • Banking and trading: Customers can use bots to get information and execute transactions.
  • Service provider: If you want to order a pizza or a car, schedule a meeting, or appeal a parking ticket, there’s a bot for that!

This last use case is potentially very disruptive.

For example, if the user wants an Uber ride, they could order a car via the Uber app, but a message is easier. They could simply instruct the bot: “Order me an Uber to go to my office at 5 P.M.”

The bot has to understand quite a bit about that command and about the user, such as: Where is the user now? Where is the office? What is the user’s Uber ID?

Now suppose the user simply said: “Order me a car to go to my office at 5 P.M.” If the user does not have a preference for Uber, Lyft, taxi or any other particular driving service but the bot provider has a deal with one of those driving services, it could route the request to the driving service they have a deal with, making itself a powerful intermediary in the transaction.

This is power.

If users can use messaging apps to satisfy many needs, this does not bode well for the apps they supplant. For example, Google can now answer many types of questions. Consider if a bot can answer those same questions, but does not show corresponding ads? Disrupting Google Search would have enormous impact on the economics of the Internet.

To bot or not to bot?

From a user perspective, what’s not to love about bots? You don’t have to discover, install or update them, learn a new user experience, or even leave your messaging app. You simply talk to them. And, they can de-clutter the mobile experience.

In reality, there can be things a user may not love. Think of a bot that keeps butting in and pushing spam. Some “intelligent” bots learn as they go, and develop some very unpleasant habits (see Microsoft’s Tay).

Practically speaking, apps and graphical user interfaces enable rich, complex user experiences that simple messaging apps cannot duplicate. If it becomes a burden to interact with the bot (“It doesn’t understand me and asks too many questions!”), users will become disillusioned quickly.

But for the right use, bots can be the perfect tool. In fact, I expect as the number of bots grow, the discovery problem faced by apps today will emerge.

In the world of finance, Thomson Reuters Eikon platform provides critical messaging and chatroom capabilities that bring together financial professionals around the globe. We are exploring the best ways to enable developers to add bots to Eikon Messenger, our secure and compliant business tool.

Getting Started: Want to build a chatbot?

If you want to build a chatbot, there are several questions you need to answer:

  • First, is a bot really the appropriate tool and does it provide the desired user experience? As noted earlier, chatbots have their limitations.
  • Next, where will this bot connect with users? The answer should be, where your users are. If you are using the bot inside your website’s chat tool, you need to use the right technology. If you want to connect with Facebook or Slack users, you need to develop for those platforms.
  • Messaging platform providers are each offering tools to help you develop bots, and there are several vendors that enable you to write once, deploy anywhere. The market is still young and there are few standards to guide you.
  • How smart does your bot need to be? It’s easy to build a simple bot that can recognize keywords and return static answers (like an FAQ). In general, chatbot technology has not advanced tremendously and often depends on hand-crafted rules.

Machine learning is starting to be used by some tool providers, but there is no out-of-the-box artificial intelligence; you get out what you put in. The good news is that natural language processing has advanced so that becomes less of an issue. Expect to spend considerable effort to build a sophisticated, domain-specific bot.

Chatbots are the oldest newest game in town, but if played well, can yield significant rewards.


Learn more

Visit Innovation @ ThomsonReuters.com to learn how we bring together smart data and human expertise to find trusted answers.

  • Facebook
  • Twitter
  • Linkedin
  • Google+
  • Email

More answers