Don’t Build a Bot Unless it Replaces a Human

Too many damn bots, doing too many strange things

I’m just a tiny bit late in saying this, but bots got really, really big. To quote one of my favorite blogs —


Don’t get me wrong. I’m a huge fan of the impending wave of intelligent agents. They’re going to become the magical servants humanity has always deserved. Just ask Dennis R. Mortensen and his team at I harassed them digitally and in-person for two years so DCM Ventures could give them our money. No one can deny that there is a lot of cool stuff happening at the intersection of machine learning and text-based interfaces.

I love that powerful tools like TensorFlow, Torch, and Watson are now at the disposable of anyone with an internet connection. I love that most of the world’s messaging platforms are looking more like web browsers than communication channels. I love that now, we actually have machines who are indistinguishable from humans. But, that’s where my hang up is. I believe that the purpose of a bot should be to replace a human, not another application.

I’m no HCI specialist nor am I a UX designer. However, it seems that in our mad rush into bots and conversational UIs, we’ve made a very bold assumption that consumers want a conversational UI. But is it a good UI? Take the example of shopping — is it better to browse a full web page of SKUs or browse things one by one via FB Messenger? Even the CEO of Spring, FB Messenger’s key conversational commerce launch partner doesn’t think that their integration will amount to much. Here’s a quote from The Information:

But Spring CEO Alan Tisch said he had “very low expectations” going into the deal with Messenger to operate a chatbot for shopping. It was “great for brand alignment and PR but it’s a huge shift in behavior you’re asking from users,” he said in an interview with The Information. Messenger’s “not about to replace shopping.

I’m betting that Messaging/Conversational Commerce/Invisible Software or whatever you want to call it won’t replace apps. Dan Grover, one of the best thinkers in messaging put it best:

“Bots won’t replace apps. Better apps will replace apps.”

To understand why, we have to look at the reasons we build bots:

1. Bots don’t make mistakes (unless the human screws up or your training data is bad. garbage in, garbage out)
2. Bots work 24/7
3. Bots have no GUI and can seamlessly fit into existing workflows such as Slack or E-mail
4. Bots can take advantage of existing channels of distribution i.e. messaging platforms

The problem that I see with a lot of bots is that they definitely check reasons #1 and #2, but do nothing for #3 and #4. That is, they force users into new habits all for the sake of being text based/a bot. Even worse, some have built separate text based apps with no integrations into the dominant messaging platforms or communication protocols (I see a giant line item for CAC in your future). In other words, a simple app or web service could do the job.

I’m not a bot hater (even if I were, I wouldn’t admit it because I’d be banished from the valley). Again, I think bots are great IF they replace humans. To replace humans you need to:

  1. Have built automated intelligent solutions based on machine learning and not simple decision trees/text templates (humans are complex creatures and it takes a lot of training to mimic them)
  2. Insert yourselves into the usual workflow of said human worker that you’re replacing
  3. Not introduce the user into a worse experience

Again, I’m not a bot hater. If anything, I’m rooting for the bots so I guess that makes me a human hater. Don’t build bots to replace apps, build bots to replace humans.

P.S. — To those of you who are screaming Skynet. See Luddite :P

But in all seriousness, the maturation of artificial intelligence as with the introduction of any technology will produce more jobs than it “destroys” on a net basis. 100 years from now, we’ll look back at all of the repetitive, low level jobs humans currently do and think to ourselves— I can’t believe people actually used to do that.

That’s a topic for another day!

Thank you to Matthew and Tim Chen for being my thought-partners-in-crime!

Enjoyed that read? Click the ❤ below to recommend it to other interested readers!