The art of conversation in the age of bots

4 min read
Patrick Stafford
  •  Jul 30, 2018
Link copied to clipboard

The majority of human history has been dominated by one-way conversations. Governments, businesses, and authorities talk to you, not with you.

But that’s changing. The dawn and ubiquity of social media, and now virtual assistants, has created an environment in which every little interaction a business makes can be broadcast and criticized. The playing field has been leveled.

This changes how we talk. According to a report from Drift, 41% of conversations conducted on websites are from executives. Naturally, chatbots and virtual assistants fit right into that ecosystem. In April 2017, Facebook announced more than 100,000 chatbots were on the platform. By 2020, Gartner predicts chatbots will power 25% of customer service interactions. Bots, Siri, and Cortana are the future.

“By 2020, Gartner predicts chatbots will power 25% of customer service interactions.”

Twitter Logo

This presents a challenge for copywriters and designers to engage more naturally with customers. According to a Forrester survey, people want their chatbots to be “polite, caring, intelligent and funny.” Hell, 12% of bots on Messenger were asked to tell a joke or amusing story.

Related: Learning UX design? Build a chatbot.

Copywriters need to learn how to write conversationally if they want to make an impact and find meaningful work in the next decade. But conversation isn’t easy to emulate.

So, where to start?

Tone is everything

Think about the way you and anyone else you know speaks. Some people are agreeable and personable, so they use lots of exclamations and “active listening” sounds (“mmhm,” “sure,” etc.). Others are more serious, and only say what needs to be said.

In the same way, your company should have a tone of voice. This seems obvious, but one study found just 20% of Fortune 100 companies interviewed had established a tone of voice. Of those, 94% said they have no intention of creating one.

“Copywriters need to learn how to write conversationally if they want to make an impact and find meaningful work in the next decade.”

Twitter Logo

One Salesforce report points out the basics: Establishing a tone of voice requires digging into how you want your brand to be perceived. Most importantly, a tone of voice influences how you talk to people; how you have conversations. “At the end of the day, brands are run by people who communicate with other people.”

So if you’re going to create a chatbot or even write web copy, you need to understand and articulate your brand’s tone of voice. If you don’t, you’ll end up sounding like everyone else. Functional? Maybe. But hardly inspiring.

The basics of conversational copy

Technology has changed the way we talk. Just like languages develop over time and through culture, internet communities have created dialects of their own. Tap into Tumblr and you’ll see posts written in capital letters, with no punctuation—and people understand it.

A 2016 piece in The New York Times examined a modern phenomenon: the disappearance of the period or full stop. As most people talk over text, that simple punctuation mark has morphed from good grammar into a passive-aggressive gesture (which Jimmy Kimmel helpfully observes).

You can’t write conversational copy without understanding how conversations happen in the 21st century.Twitter Logo They’re mostly over text in short bursts, and often defy grammatical conventions. Also: Most people wouldn’t talk to you if they had a choice.

According to the same aforementioned Drift report, 34% of respondents said they’d use chatbots as a way to get connected with a human. On top of that, 37% say they use them for quick answers, and 35% said they use them to resolve a complaint.

“If you’re going to create a chatbot or even write web copy, you need to understand and articulate your brand’s tone of voice.”

Twitter Logo

“Not all consumers are ready to abandon human-to-human interactions entirely, and some aren’t sure they trust the technology to perform certain tasks without making mistakes.”

Every word you type needs to be created in that context. Just think about it within a work scenario: If you walked up to someone who was staring down at their computer and talking in short bursts, would you ask them a complex question in a bubbly voice?

No. You’d say, “You seem to be busy, I’ll come back later.”

Or perhaps, “I’m really sorry to interrupt you, but…”

Within the context of a chatbot or assistant you don’t have the luxury of body language. So it’s safe to make the assumption that most people are talking to you because they’re not talking to a human. You’re their second choice.

7 rules for conversational writing

So with all that in mind, how do you actually write with a consistent conversational tone that speaks to your users?

We’ll look at some examples soon enough, but these principles are a good starting point:

Don’t write the way you talk
This advice is common enough, but if you look at the way you talk there are plenty of pauses and “ums.” Don’t do that. Instead, read it aloud: Does it sound stilted or awkward? If so, change it up.

Always use contractions
In the context of a bot, it can be tempting to sound a little more formal, but that can easily turn robotic, which is not conversational. People use contractions, and so should your bot.

Use conversational interjections
Instead of a full sentence, words like “yep,” “ah,” and “sure” show your user that you understand.

I don’t know about you, but when a customer service rep says something like, “I can see that you are having a problem. I’m sorry to hear that,” I don’t think they’re sorry at all.

Related: The ultimate guide to chatbots

Instead, I’d rather hear something like, “Ah, that’s no good. Let me see if I can help.”

“People use contractions, and so should your bot.”

Twitter Logo

Don’t overuse exclamation points
“Hi there! It’s great to talk to you!”

No it isn’t. Exclamation points seem fun, but they can easily become overbearing. They place the reader in a state of readiness or obligation, and might make you seem like you’re glossing over a problem.

Understand the tone
If a customer has a deep problem they’re extremely concerned about—like cancelling a flight due to the death of a loved one—it would be callous to say, “Ah, that totally blows, man!”

That phrase might be something you hear in a conversation, but it’s certainly not appropriate here.

This is where bots need to get better at targeting words and changing tone based on trigger phrases. If you detect that a user is pissed off, playfulness should be toned down and sentences should become shorter and more direct.

Write in an active voice
I’ll keep it simple: An active voice means that the subject of a sentence does the action.

Active: I throw a ball.

Passive: The ball was thrown by me.

The difference is subtle, but using the active voice implies momentum. Keep the conversation rolling.

Keep it short
Remember, conversations occur between two or more people. Don’t dominate the conversation with long copy (though long copy can be useful at the right time). Say what you need to say, then request input from the user.

For instance, let’s say someone is looking for a flight:

“Here are some options for you. What else can we do for you?”

“Say what you need to say, then request input from the user.”

Twitter Logo

You give a response, then offer more help. Responses should be short, too. Think about how your friends react to questions:

“Hey, can you pick up some beer on the way here?”

You don’t respond, “Sure thing, buddy!” You might say, “Sure,” or, “Okay.” The same rule of thumb works for conversational copy.

Be playful
Just because you need to respect tone and cadence doesn’t mean you can’t have fun at the right times. You can use slang in an ironic way, or even ask the user for a joke—but it all comes down to timing.

Conversational copy IRL

It’s hard to keep these principles in the abstract. Let’s take a look at a few chatbots that get to the point while using conversational copy well.


We love that Chatbot Magazine’s post included this as a reason why Poncho was so great—notice the declarative statement at the start, with an exclamation point.

The joke is funny, but the “cool” at the start is a nice, subtle way of letting the user know they’ve input the right message.

Also note the contractions.


Image via eBay ShopBot

The word awesome is often overused and can seem excessive, but without the exclamation point, it fits right in here.


Sephora’s use of emojis is cute, but the real kicker comes with the third question: “Do you want to…”

This wouldn’t work as well if you demanded, “Take a quiz so we can provide you with tailored recommendations.” Instead, it’s polite and inviting. Like an actual conversation.

Insomnobot 3000

Weird name. Good copy.

This tool is meant to be a companion for people who can’t sleep, and you can’t even tell it’s a bot. It’s not afraid to be short, to the point, and a little insulting. Words like “Yeah” and questions like, “Maybe I should call in sick to work?” sound exactly like a human.

Keeping your users happy

Let’s be real: Conversational copy is not the only reason your chatbot is going to convert. At the end of the day, people have a problem and they want it solved. But it can delight your users and make them feel as though they’re being taken care of.

Which—in a world soon to be dominated by bots—is just as important.

Collaborate in real time on a digital whiteboard