Enterprise Messaging News + Insights by Mio

Conversational UX — What B2B and B2C Bots Can Learn from Each Other

How We Chat April 28, 2017


Conversational UX — What B2B and B2C Bots Can Learn from Each Other

After speaking with many bot creators last month, we know nobody’s bot is perfect. Here are a few things I’ve noticed business and consumer bots could learn from one another.

What B2B Bots Can Learn from B2C Bots:

1. Infuse Personality. Many of you B2B bots see your bot as as a utility — which is understandable, if you’re coming from an integration. That’s great for sticking to a clear use case (something we all know is key for bot success), but doesn’t mean your integration bot interactions have to be rigid.

I’m not proposing that you hire an intern to infuse GIFs and emojis into every interaction (unless that really is your brand). But, infusing personality in few key places can lead to enjoyment, and therefore, retention.

For example, slash command flows are a specific area to infuse personality that could deliver real results. If you’re already successfully using slash commands, you probably fall into one of these camps:

  • You’re lucky! Users are adopting your slash commands because they’ve used them before. Maybe it “just works” or maybe you’re riding an initial wave of novelty and power-user traction. Only time will tell!
  • You’re using science! You recognize it’s hard for some people to remember to use slash commands, and took a page from Hooked (whether you realized it or not).
From "Hooked" by Nir Eyal

Here are two examples.

  • It’s a known fact that sales folks have a love-hate relationship with Salesforce. The Troops Slack experience drives use of Salesforce data entry by rewarding their slash command actions, like changing a deal state to win with a different “gong” celebration GIF each time.
  • Taking the time out of your busy workday to show appreciation towards a co-worker isn’t something most of us are wired to do, but Growbot lets you invoke “props” with a command line. The action to take is simple and memorable, and the reward is seeing everyone else chime in too. Repeat a few times and you’re on your way to team culture, all from a little slash command.

So, infuse personality (like most B2C bots already do!) when designing slash command flows, and you might just create a winning user habit. They’ll forget they’re actually doing work!

2. Be contextual within the platform. Messenger bots have already discovered the power of tapping in to contextual information. Miguel, founder of Claimbot, says he “use[s] the data from Facebook and call centers to replace filling out forms.”

Some users are surprised at first, but they generally they want the least path of resistance, which is a great reason to use a bot over a website. The more your users interact with your bot, the more context you have to make the experience better and smarter.

For bots in the workplace, this simple concept should spawn ideas. There’s a plethora of contextual information available to take advantage of (within privacy and contractual boundaries, of course). A job title, time zone, national holidays, “last seen online,” or repeated behavior patterns like when your user typically engages with your bot are all examples of context.

3. Test conversational-style experiences in small doses. Consumer trends on Google search queries and adoption of voice interfaces like Alexa already highlight that we’re searching for information in a conversational method. Not to mention the new “chat as a landing page” trend that we’re likely to see more of.

Yet, we all know now conversational style has proven to be harder than expected, and enterprise bots may be weary of using third party NLP tools, so they’re more likely to stick with command-style interactions.

But, are your slash commands core to user experience (like the examples above), or just leftover from older app integration approaches from previous years? Now may be the time to experiment with a subset of your bot flow that lends itself to a basic question/answer format using keywords.

Luis Borges Quina, CEO of Ottspot noticed this approach would benefit his users during setup and onboarding education. “Yeah, NLP was hard, but we decided to go from slash commands to keywords because it was easier for users. Now we want to have answers to common usage questions like ‘things you can do with Ottspot’ inside Slack so people don’t have to go to the website to find that information.”

Think about how you can conversationally help your users get to the information in your content database faster now, and reap the benefits of having this foundation later. The time to organize your content data for conversational lookup is now.

What B2C Bots Can Learn from B2B Bots:

1. Narrow in on a specific, valuable purpose for open-ended NLP. If you’ve read any “how to build a chatbot” articles, you’ll find they mention use cases.

But, when speaking to numerous bot creators, I didn’t hear a single consumer bot mentioned when I asked what bots they regularly use personally.


Perhaps B2C bots need to take a page from the utility-oriented B2B bots that are all about that GSD life. Sure, maybe your bot is more “play” than work, but you should avoid being too general, like a website is, or trying to promise too much functionality.

Give your users a new reason to come experience your bot, like getting a personalized consultation, customer support, or scheduling an appointment.

You may have to test your chat use case a few times to get it right, so don’t get too attached before testing.

Be wary of creating a fully conversational “search” use case unless your content library is fixed. Searching within a broad and undefined category will likely lead to headaches and frustrated users who are accustomed to GUI-based searching with filter tools.

“In certain use cases, a GUI simply works better than chat,” says Hussein Fazal, CEO at SnapTravel. “If you know exactly which hotel you are looking for, NLP can work. However, if you are deciding which hotel to stay at, you need filters, lists, and maps in order to get a complete picture and make a decision.”

B2C bots should also consider how a first time user experience may be different than a repeater experience. It’s going to take time for the general public to become comfortable with bots, so you should assume your bot might be the first bot they’ve ever used (no pressure!!)

Consider a hybrid design approach, like Kristi and Chris Colleran from Sciens.io, creators of Eventbot. “Using both guided visual tree flows, as well as NLP tools like LUIS and API.ai give people the guidance they need at first, then once they get a feel for it, they want to text. You have to be able to handle both, or it will be a disappointing experience,” said Kristi Colleran who serves as the bot’s conversation designer.

2. Reduce decision fatigue to get started. B2B services love the free trial concept. Managers want to try out a new tool before investing their budget. For consumer bots, users often just want to try out the service to find out if its something they want to continue to investing their time on.

Quickly getting to the “wow” moment without too much effort is important — but many many B2C bots try to take a new user through a series of questions to show how powerfully customizable they are.

Dan Manian, Co-founder at Donut.ai noticed they had more successful onboarding and use of their product when they adopted this approach. “At first we asked what day of the week they wanted to schedule pairings, what time, and the frequency…but then we decided to constrain some of those options so that people have fewer decisions to make.”

Set a reasonable default way to get started, and let your users change those defaults later when they get a general feel for how your service works.

3. Create notification preferences. B2B bots are sensitive to the noise they create in workplace chat platforms, as employees easily will tune out notifications that are deemed non-critical.

Cory Shaw from Gofaster.io, a daily website performance monitoring tool for Slack, noticed that while automated daily roll-ups were the reason customers signed up for the service, many wanted immediate alerts. “We are thinking about the mechanics of triggers for notification, like having the user tell us when x or y threshold is hit, send a message, or saying that we noticed that image size went up and asking ‘do you want more alerts like this?’”

B2C bots should test in a similar way, not unlike how mobile apps approach push notification testing.

Figure out what proactive messages lead to increased engagement and are not seen as spam. Messenger bots flood me with messages all the time, as if they don’t know I can turn off all messages from them.

Muting Messenger conversations or blocking an SMS number is the new “app uninstall,” and irritated users will do the work to find it — so be useful, and test!

You shouldn’t expect all users to use your bot at the same frequency, so take a cue from their engagement levels from time to time. A simple “Hey — you seem quiet lately. How often should I reach out?” helps reduce frustration.

Whether you’ve built a business or consumer-oriented bot, you should thinking about your existence as a business beyond one platform….which is exactly what Message.io can help with! We’re currently in private beta for expanding your bot’s reach across platforms. If you’d like to get on the list, sign up here!

Have any tips on best practices for B2C or B2B bots? Tell us on Twitter!

Leave a comment

Leave a reply