Travis Nelson
Mar 20 · 4 min read

So you want to build a chatbot? To get there, you can take a lot of paths with varying levels of complexity (and varying levels of results).

Building a bot is building a product. Approaching chat, whether it is text, voice, or part of a multi-modal experience, should be done in the same way you plan out other products. You need to have consideration for business goals, user goals, success metrics, and resource planning.

We will take a high-level overview of what product planning can mean for chatbots, which is essentially asking yourself a lot of questions.

(Note: there are a lot of reasons to build a bot, but I’ll be focusing on commercial chatbots here.)

Hold up: why are we doing this? (source photos by Benjamin Davies and Rock’n Roll Monkey)

Why are you building a chatbot?

Why you are building a chatbot requires research, and the amount of quality research you do will be correlated with the success of your bot. It is possible to get lucky and have a successful chatbot by taking a shot in the dark, but as anyone who runs a product will tell you, taking the time to understand why and how you are doing something increases the probability that you will obtain market fit.

You should be able to answer the following questions:

  • Is chat or voice interaction the best solution for your users?
  • What does the bot need to be be able to do?
  • Who are your users that will interact with your bot?
Find the right type of home. (source photos by Luke Stackpoole and Sorry imKirk)

Where will your bot live?

The first part of where will your bot live refers to your users: where do your customers want to interact with your bot? Are they going to chat with your bot in their existing social channels, or are they seeking you out somewhere else?

The second part refers to the channel itself: what are the capabilities required by your bot and can the channel meet them? If you have requirements that are not specifically supported by your channel(s), like getting location data, can you work around that by asking for an address or intersection? Will you need different capabilities and workarounds for different channels?

The third, and final, part is: where is your bot hosted, and does it have access/integrations into the channels you want? Depending on your goals, you may want to select an extensible platform with a lot of available channels to grow your bot, or an available API to build out future integrations, even if those channels are not natively supported.

Who is your bot?

Your bot will be an extension of your brand, but you can also make room to define a separate experience from your primary branding voice by using a character or personifying your bot. Deciding who your bot represents will also affect if or how the bot refers to itself. For example, your bot could say “I am looking for that info” if the bot has its own personality, “We are accessing that information” to represent your company, or omit pronouns altogether “Pulling up profile information”.

People interacting with a bot will treat your chatbot like a bot most of the time, but conversational conventions will still come into play and can improve your customer experience.

Deciding who your bot is will likely include your branding or marketing teams.

Success will take a while. (source photos by Wolfgang Lutz and Kaboompics)

What is success?

As with any product or feature, you should define what success looks like for your bot. What metrics do you need to capture and meet? Common examples include: turns-per-conversation (higher in brand engagement use cases; lower in a customer service or FAQ scenario), click-through-rates, conversions (like sign-ups, survey completion, or product purchase), live agent escalation, and customer satisfaction. Key performance indicators (KPIs) will vary depending on the use case and overarching project goals, but regardless, these metrics should close the loop on why you are building your bot.

Pre-defined metrics are a solid place to start, but you should also be prepared to revise your targets as your bot is out in the wild. You may notice emergent behaviors with your users you want to capitalize on. Because messaging and voice are relatively new engagement platforms, you may learn things about your customers that didn’t surface in any other user research, as they will interact differently and have different expectations in a new medium.

How do we achieve success?

At this point, you know why you are building a bot, where to build and deploy it, who your bot will be, and what to measure. But how do make all of those things happen?

Going back to product planning, you are going to need resources to execute. Depending on your ambition, your resources can go from just you to small teams to entire companies to looping in outside agencies, and this is all before thinking about the launch plan and marketing.

Plan to have resources to update your bot on a regular basis. Even if you intend to have the capabilities remain static, you will still need someone to manage new inputs from users that haven’t been defined yet.

The key to a successful bot comes after initial deployment, when you find out exactly how your users are interacting through their conversational inputs.

If you’ve decided that a chatbot is right for you, build one for free on Pandorabots. You can also read more about chatbots by checking out our other blog posts.

pandorabots-blog

The leading platform for building and deploying chatbots.

Thanks to Pandorabots

Travis Nelson

Written by

Head of Product and Design at Pandorabots

pandorabots-blog

The leading platform for building and deploying chatbots.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade