All hail the hype cycle — 2016 was nothing but acceleration for the burgeoning bot economy. We experienced Gartner’s “innovation trigger” and reached the “peak of inflated expectations” in under 10 months. Congratulations, everyone!

Now comes the sobering reality that companies — actual, living, breathing customers — need to be able to implement these things and demonstrate real business value. While some may long for those magical days of canned onstage demos and offensive rants from Microsoft’s Tay, there’s no escaping the reality that bots are just a fad if they don’t produce for the brands that build and buy them.

Unfortunately, because of all the hype generated in the past year, many brands have been suckered into biting off more than they can chew. Sales of AI technologies are going through the roof, but value-generating implementations lag well behind. For the $8 billion worth of revenue generated by AI companies in 2016, I find it hard to believe those buyers are on track to see more than $8 billion in ROI.

It’s not that the technology they purchased is necessarily bad, but customers often bought too much and tried to do too much from the get-go. I saw this in virtually every sales meeting I had in January — mountains of software and little notion of where to begin. Here’s what I tell these companies when I find them in this predicament.

VB Transform 2020 Online - July 15-17. Join leading AI executives: Register for the free livestream.

1. Narrow is the new smart

I’m amazed by the grandiose scope of the automation projects businesses envision. Some want to replace an entire contact center in under 12 months. Others want a bot that can handle technical support and conversational commerce across a 500-product portfolio through multiple sub-brands and multiple conversation mediums.

These broad ambitions are destined for failure in almost every case. Bots aren’t all that complicated on their own — they’re just an interface with the customer and in most cases should act as the delivery mechanism for conversational intelligence. But the brain behind them needs to extract meaning and take the appropriate action. More complex customer inquiries often require integrations with a bunch of legacy back-end systems, which adds another layer of complexity that companies often overlook. It’s the brains, not the bot, that brings real value for brands.

If you are even able to get such a fully general purpose bot or automated experience to production in the next two years, I’d be stunned if it didn’t disappoint customers in nine out of 10 interactions, because the brain needs time in production to be trained and honed with the proper data, question sets, hand-off mediums, and voice.

A narrow approach — with support for a finite set of customer inquiries and limited but compelling functionality designed to make customer’s lives better or more efficient — can now deliver great experiences 90 percent of the time. And most businesses can get the first set of use cases live in a matter of months not years. Pick a narrow set use case for your first project, limit the functionality to what’s essential and going to make the most impact for the brand and end customer, and pour the rest of your efforts into ensuring a quality experience on those fronts in the various chat and voice channels.

Other than succeeding, what benefits do you get from a narrow approach to bots and automation?

Data, data, and more data. Once you’re live, you can collect invaluable data about the questions your customers ask, the features they want to see next, and so much more about their needs as they evolve over time. That leads me to my second piece of advice…

2. Make iterative great again

It wasn’t too long ago that “iterative” was among the most cherished buzzwords in the tech sector. For some reason, we’ve forgotten the lessons we learned about iterative development and walking before you run.

Today’s top brands are obsessed with omnichannel everything. In the case of bots and automated experiences, which primarily manifest within chat apps today, brands want to know that they can go live on every channel their customers use on day one.

Here’s the catch: if your bot delivers a bad experience on Facebook, it’s going to deliver a crappy experience on Twitter. If you want to ensure success with your bot experience, don’t get greedy about channels. Get it live on the most valuable channel to your customers and keep improving on your conversation models until you’re confident you can perform well in more channels so that you can ensure that you then make the most of the unique properties and functionality of each channel.

It’s so important to resist tunnel vision when you hear stats about millennials preferring chat and self-service. While it’s true that well over half of millennials prefer 2-way text chat for customer service, 100 percent of customers hate bad service experiences. The medium matters, but only if the experience meets or exceeds expectations.

Build something that works. Then iterate and scale to new channels.

3. If a bot falls in the woods…

My final piece of advice seems like a no-brainer, but almost everyone screws this one up. Bots, despite all the hype, do not market themselves. If you build it, there’s no guarantee that anyone will notice.

Your customers will absolutely appreciate and adopt a high-quality bot. They can generate revenue and they can drive customer satisfaction. They can do a ton, if you show your customers where they are and how to use these automated experiences.

The exact nature of the promotion you need to do is highly subjective to your business and your customer base. But the one thing you can do to de-risk your automation investment from the start is to get your marketing team started on a rollout strategy. Even if you get a few hundred users with promotion, you’re not getting a statistically useful amount of data from those interactions.

Start your marketing plan in parallel with development of your first bot project. Have the teams work in lock-step along the way, and don’t underfund the effort.

Succeeding with bots is not rocket science. A little common sense, a walk-before-you-run approach and some basic communication can get you from theory to production in less time, at dramatically lower costs, with tangible results to show for it that you can continue to build on and expand into more AI-driven experiences over time. That’s the value proposition bots were supposed to have all along, and it’s there for the taking if you have the discipline to capture it one step at a time with the right conversational intelligence behind it.