Planning A Chatbot
- Abbey Brown
Quick links in this article:
In this article, we’ll explore the definition of a chatbot, the channels and use cases, and steps for designing your own chatbot journey.
Prefer to take this offline? Download our chatbot planning guide here:
Chatbots vs Quick Replies
Put simply: Quick Replies are a feature of chatbots and automated conversations.
Chatbots are simple (yet effective) rule-based flows, using if-this-then-that logic to interpret what a customer is saying and offer a response. However, the open-text nature of chatbots leaves space for complexity - such as a typo you haven’t accounted for in your logic. Chatbots without quick replies are more complex to create, and more limited in their applications.
Quick Replies reduce that margin for human error by offering customers a selection of buttons for their reply. These buttons have hard-coded values that you can use to make more intelligent decisions in your chatbot journey. That means you can go further, and do a lot more, before passing the conversation to a human agent.
In the example below, you can see how Quick Reply buttons can be used in combination with open text responses.
Channel Availability
Chatbots started in webchat - just like the example shown above! But they can now be built in a range of digital channels, including: Webchat, Facebook Messenger, Instagram DMs and WhatsApp.
Chatbot journeys are essentially the same, no matter the channel - though the config steps in Gnatta may be slightly different. In these articles, we’ll describe the steps for setting up a chatbot via Chat.
If you’d like to set up a chatbot in another channel, such as WhatsApp, please get in touch! We’d be glad to get you started: Help Centre Home
In the meantime, feel free to check out our WhatsApp Chatbot demo journey: FuelsUpp
Chatbot Types
There are three major approaches for chatbots, regardless of the channel they’re implemented on. If the channel is one you’d typically resource with live agents, we’d strongly recommend Self-Serve (Live Fallback)!
Before proceeding, you’ll need to decide which type you’re going to be planning for.
Self-Serve Bot
Completely ‘offline' - no agents needed to man your channel!
Used to resolve simple queries, like FAQs
Escalation flows can be used to flag high risk interactions for follow-up
Can be available to customers 24/7
Self-Serve (Email Fallback)
If the chatbot can’t resolve it, send a transcript to email for your team to pick up
Reduces frustration vs a fully unmanned solution
Doesn’t need to be staffed as a live channel
Self-Serve (Live Fallback)
If the chatbot can’t resolve it, transfer directly to a live agent
Out of hours, send a transcript to email for your team to pick up
More effective than pre-chat surveys on their own, with a good chance of deflection
Best-of-both in terms of CX and operational efficiency
Features to consider
Whilst you’re planning your chatbot journey, consider whether you’re making best use of these popular chatbot features:
Quick Replies - Use buttons to navigate chatbot journeys instead of free-text so you can reach deeper answers
Anything Else? - Instead of ending the chat when the question is answered, consider returning to the main menu of options
Integrations - Collect external data (like an order status) and present that back to the customer in your chatbot journey
Surveys - Planning a journey for webchat? Don’t forget your pre-chat and post-chat surveys can handle run of the mill questions too - like Customer Name, Email and so on.
Planning
Whilst all the nuts and bolts will be happening in Gnatta, sometimes you just want a pen and a pad of paper to get stuff like this done. We get it. Here are some free templates drawn in popular whiteboard-style software options to get you started:
Canva: Chatbot Template
Figma: Try Figma Template
LucidChart: Try LucidChart Template
Fancy whiteboarding software not your style? No worries. Feel free to draw them up wherever, and however you like. Just remember, you’ll need:
The exact written copy you’d like to use in your journey
The full URLs for any links you’re using
Visible logic & connections on each pathway
A clear ‘end’ action on each pathway – end chat, pass to agent, or an email follow-up?
Notes for how you’d like to approach error-handling
Scheduling concerns – is your journey different in-hours?
Split your journey maps into sub-paths to keep things clear
Finishing your pathways
Finishing a chatbot path should do one of three things:
Connect an agent
Send an email transcript into your queues
Close the chat
Be sure to note down what you’d like to do at the end of each path.
Building your chatbot
Plan signed off by the team and ready to build? Let’s get to it!