Sixty percent of people think chatbots will make companies feel less human. But industry insiders disagree. Back in January 2016, Chris Messina (tech guru and inventor of the #hashtag) predicted:

Computer-driven bots will become more human-feeling, to the point where the user can’t detect the difference, and will interact with either human agent or computer bot in roughly the same interaction paradigm.

His point is that chatbots will eventually become so slick at communicating that they’ll sound just like humans (which is funny, since saying things like ‘interaction paradigm’ makes him sound more than a little robotic). Irony aside, bot designers do generally agree that that’s the destination, and many go out of their way to make sure that people know they’re talking to a bot. They don’t want a scenario where someone thinks they’re chatting away to a human, only to find they’ve been communing with an algorithm all along.

But I think this is the wrong debate. I’ve been chatting to a bunch of robots recently, and they’re already better communicators than a lot of real people in business. The danger isn’t that bots don’t measure up, or that we can’t tell them apart from real people. It’s that chatbots’ conversational skills leave their human colleagues behind. Bots will end up being the better writers, and people will sound like robots.

Compare these two chats I had recently, one with a bot, and one with a human agent. Can you tell which was which? Which one strikes you as more natural and conversational?

The chat with a real person is the one on the right. It’s more formal, slightly awkward, and they weren’t very good at answering my question – which isn’t ideal considering what business they’re in.

When humans write at work, we often sound anything but natural. Business writing is famously cold, confusing and packed with incomprehensible jargon – and live chat agents are no exception. The days of hilariously bad live chat are (generally) past, but the fact that over 70 percent of people would rather solve a problem without having to talk to an agent at all suggests companies still aren’t getting it right.

The Bots Are Catching Us Up, Fast

At the moment, bots are pretty limited in what they can talk to you about. Most of them force you to pick from multiple options to keep the conversation moving (like the one above), and even the ones that can respond to actual typed questions can only give you pre-scripted answers. (Although experts are predicting that it won’t be long before they can give real answers that adapt to context.)

If your medium is a conversation then you better sound conversational But even with those limitations, the bots are pretty good at sounding human. Chatbot pioneers are fully tuned in to the need for bots to sound natural, because they understand that if your medium is a conversation then you better sound conversational. For that reason, lots of software designers start with the ‘personality’ of their bot before they even begin to write code. And they involve language experts at the scripting stage, who really understand the nuances of how conversation works.

For brands with more distinctive personalities, that’s a bigger challenge. It’s not just about being natural. Their bots’ personalities have to reflect the quirks of the brand. We’ve been banging on for years about the importance of baking your tone of voice into everything you do, and forward-thinking companies are turning their bots into ambassadors – like DOM, the Domino’s Pizza Bot, who I spoke with the other day.

DOM sounds like Domino’s. He’s a bit irreverent, and makes pop culture references. (It does feel like Domino’s have missed a trick here though – it wouldn’t have taken much to give DOM a few things to say about his favourite pizza.)

These early adopters are showing how to balance the need for really effective communication with reflecting what the brand stands for. But as bots become cheaper and more common, companies have to be careful not to cut corners. People chat with chatbots – so clunky language won’t cut it.

Good Chatbots Need Better Humans

As bots deal with more and more everyday customer questions, your real live humans are going to be left to deal with the difficult ones. Only 29 percent of people think a bot will be able to handle complex problems, and no-one expects chatbots to take on everything (for the foreseeable future, anyway).

It’s pretty unlikely that someone who jumps from a chatbot to a human agent will do it because they’re getting everything they need from the conversation already. They’ll be trying to deal with knotty technical problems; they’ll be asking difficult questions; they’ll be frustrated – or just plain angry. That’s why bot designers recommend always keeping the option to switch from bot to human chat open.

In those moments of truth, when an already-frustrated customer gets shunted from a bot to a live agent, your people have to be great writers. Those high-pressure situations determine whether customers stick with you or run a mile. Some studies say as many as 91 percent of customers who’ve a bad experience won’t come back. But it’s when we’re dealing with complicated problems or giving bad news that our language tends to become stiffer and more awkward.

If you’ve got technical experts dealing with trickier customer questions, ask yourself: have you really given them the skills to blend their knowledge with empathy, and to write in a way that supports your values? Are they suffering from the curse of knowledge? Customers are more demanding than ever. So if you’re investing in a fancy chatbot, don’t neglect your humans. Their jobs are about to get much harder.