Why chatbot writers need to focus more on questions than answers

Writing for chatbots requires a deep understanding of user intent. Here’s what intent is, why it’s important, and how to get it right.

A screenshot of computer code.

When you think about writing for chatbots, you probably imagine crafting a character, and their responses to different customer queries. That’s certainly a part of it, and something we’ve covered in our previous blogs on key aspects of writing for the new era of conversational interfaces.

But one of the more significant parts of chatbot copy is actually writing for the customers rather than for the bot itself, as I found out when attending a recent event on chatbot development.

Razzle dazzle

One of our clients, Lola Tech, recently held an event to discuss the launch of their new virtual assistant, Dazzle. I was lucky enough to attend, and got the chance to talk with many of the talented techies behind the virtual assistant.

A voice-activated digital concierge for the hospitality sector, Dazzle recently won an accelerator program with Marriott hotels. It was deployed in several rooms at the Marriott in London as part of a trial program.

The team that made Dazzle, and the Marriott trial, included a range of job titles, from technical staff and developers, to test engineers and customer experience consultants. One of the more surprising inclusion on team Dazzle was a corporate poet who was on hand to script the AI.

You’d think the poet was mostly responsible for tweaking the responses Dazzle gave to hotel guests. You’d be wrong. Their job also had a heavy focus on pre-empting the many ways guests would interact with the digital concierge. They didn’t write answers – it was the questions that needed a writer’s touch.

The many faces of intent

Chatbots and their developers don’t really care about what questions users are asking. What they need to know is the intent – the thing that a user actually wants the bot to do.

The problem is, they’re often not the same thing. There are many ways to show intent, and some of them – while obvious to the human ear – will absolutely baffle a machine.

If I head round to a friend’s house and ask: “Are there any beers in the fridge?” I’m not doing it out of concern that my friend is running out of drinks. I’m really just trying to ask if I can have one.

But I could ask for that in any number of ways:

  • “What beers do you have in?”
  • “Can I have one of your beers?”
  • “Gee, I’m really thirsty all of a sudden” (hint, hint)
  • “I could murder for a beer right now”
  • [Sees a beer advert on TV] “I could go for one of those”
  • “I’ve been here for 3 hours and you haven’t offered me a drink. Why are we still friends?”

There are many ways to ask for (or very unsubtly hint that you want) something, and chatbot developers need to consider every single possibility as they develop their voice and text interface tools.

If a chatbot doesn’t have properly considered intent coded in, it can result in a lot of the “I didn’t understand the question” responses. (Or at the very least, some detailed but pointless answers itemising the drink content of your fridge.)

This is a fast track to frustrating users, and putting your bot firmly into the ‘will not use’ pile.

Who writes intent?

Intent is an important thing to get right, and somebody has to come up with all the potential ways a user might show it. But who should be responsible for this?

Seeing as it’s a task that requires an intimate knowledge of language, I’d argue it’s something that shouldn’t be added to the already huge list of tasks on developers’ plates. Their skills are better applied elsewhere.

Not all teams are lucky enough to have a corporate poet on board though, or in the case of Microsoft’s Cortana team, a whole slew of scriptwriters, authors and poets to craft its virtual assistant. Even if you don’t have these kinds of writers on hand, a copywriter might be able to help – but they might need a different set of skills to what you usually expect a copywriter to have.

Copywriters writing intent questions for chatbot will need to abandon their usual focus on brand voice and tone, and double-down on their efforts to get into the minds of their audience. Only by understanding their audience (and all the varied ways they speak), will copywriters have a chance to come up with the full range of intent questions chatbots will need to respond to.

Getting into your audience’s head is important for copywriting in general of course, but for intent questions and chatbot writing it’s a truly essential skill. If you’d like a refresher, you might find this video helpful. Or you can check out this post for a more detailed look into creating audience personas.


More posts you might like…

B2B writing quality checklist

Reviewing B2B copywriting? Steal our 15-point quality checklist

Here's a secret: writing B2B content is hard... and reviewing it is even harder. So to help, we score every draft on a 15-point scale. And we're happy for you to make it your own.

Make your writing more effective

Get copywriting tips and advice — direct to your inbox every month: