Bots are showing themselves to be one of the darlings of adtech in 2016. They may potentially disrupt brand-consumer interaction as we know it, creating both opportunity and chaos for marketers.

And yet, as with websites and apps, building an effective brand asset doesn’t happen overnight — or come without challenges. And that’s certainly true for bots, whose technology is still very imperfect and will require extensive development before all is said and done. Even so, the benefits outweigh the risks, and brands should act now, as first movers will likely see major advantages.

Previously, I spoke with more than a dozen executives to learn what bot implementations hold the most promise (See: 10 ways bots can surprise and delight your customers). Here, I share the specific challenges bots present and how brands can successfully overcome them.

What challenges do bots present?

As we saw with Microsoft’s Tay — an artificial intelligence-powered chatbot that sent racist tweets — bots are not without flaws.

“At this time, I think the risks outweigh the benefits,” concluded Eddie Francis, research director at market research agency Alter Agents, although he noted bots’ capacity to answer customers quickly and efficiently and their growing ability to understand what customers need and when they need it. In his opinion, because bots are still in the development phase, they currently pose too many risks for brands.“It takes a long time to build your brand equity — having bots speak for you unfiltered is too risky for established brands,” Francis said.

Naturally, not everyone agrees. Beerud Sheth, CEO of smart messaging platform Gupshup, said, “It makes sense for brands and marketers to use bots at every customer touch point. Brands have to go where the users are — in their messaging app.”

Three concerns marketers may face

All that work on our app …for nothing?!

Adam Cohen-Aslatei, senior director of marketing at advertising company Jun Group, said brands that implement bots have to be comfortable with not only seeing a decline in usage and engagement with their own mobile apps, but also with ceding control to yet another app — messaging platforms like Microsoft, Facebook Messenger or Kik, on which their bots reside.

And this will cause frustrations for some brands and marketers. “It took the better part of a decade for marketers and ecommerce companies to master the art of the mobile app,” said Robert Tercek, a self-described “business futurist” and the author of “Vaporized: Solid Strategies for Success in a Dematerialized World.” “Everything they gained could be washed away, and they will need to start over again with a new round of customer acquisition, mastering the art of discovery and findability in a brand-new environment inside a mobile messaging app.”

Bots are a work in progress

Pypestream chatbots can help resolve many customer service issues.

Above: Pypestream chatbots can help resolve many customer service issues.

Image Credit: Pypestream

Richard Smullen, CEO of mobile messaging app Pypestream, sees peril when it comes to using bots for customer service. For instance, when a consumer has a genuine issue, there’s the risk that the bot may simply respond, “Thank you. Have a nice day.”

“So you want a bot that has a level of A.I. that is going to require investments,” Smullen said. “If brands are going to do messaging…then they should have bots, but they should know that the response is not going to be all-encompassing and won’t take every kind of inbound question.”

Smullen pointed to his own recent interaction with an Amazon bot by way of example.

“I messaged the Amazon bot and said, ‘My package is late,’ and the response from their bot was, ‘I’m sorry you’re up so late – why don’t you choose one of these movies to kill time?’ So I said, ‘No, my package is late,’ and it said, ‘We’re sorry to hear that, but if you’re up so late, why don’t you watch a movie?’” Smullen said. “That bot bombed. It was not programmed well. It didn’t do natural language processing well…I think brands need to understand that the advanced nature and the spectrum against which this bot is able to understand, respond, and interact is important, and the greater the advanced nature of this bot, the more expensive it will be and the more sentiment analysis will be required.”

Even with the promise of A.I., Francis said that anytime a bot is learning on the fly and using language, there’s a major risk of offending consumers. And any instance could easily go viral. Natural language processing (NLP) is far from perfect.

“Bots know literal meanings of words, but oftentimes humans create new slang to give words updated meanings which may not be appropriate for certain circumstances,” Francis added. “It’s not easy for bots to re-calibrate their language set on the fly, so it’s definitely risky for brands to be associated with these unknowns, especially when we live in such a curated-heavy world.”

You may not have the data you seek

What’s more, messenger apps or platforms like Facebook Messenger and Kik are walled gardens. Just as mobile app analytics followed the explosion of iPhone apps, robust data analytics for bots are coming soon. But for now, according to Cohen-Aslatei, brands may have a difficult time accessing the data they need from these platforms in order to understand their customers and provide them value.

Despite these challenges, many of the executives with whom I spoke declared that the time to implement bots is now. So…

11 Strategies for success with bots

1. Let go a little

Brands should not be scared of messaging platforms, because they can’t control the message at all times. The benefits of using bots are so much greater than the drawbacks, according to Pypestream’s Smullen. For instance, bots are always on, and so can be always available for customer service, commerce, and so on.

2. Start early

Brands that haven’t started working on bots are already behind. “The web and app paradigms demonstrated substantial first-mover advantages to the early movers. So will bots,” Gupshup’s Sheth said. “Getting started sooner will enable brands and marketers to learn, iterate, and innovate faster.”

Ben Kosinski, head of the Collaboratory, a unit within marketing agency iCrossing, agreed. He noted that while it depends on the brand and the category, being early generally provides unique opportunities to become a leader in a space – particularly among those that can execute well and then go on to leapfrog the competition.

3. Keep it simple — and focused

At this point in the technology’s development, bots don’t need to be overly complicated.

This image shows a Screenshot of H&M retail bot on Kik

Above: Screenshot of H&M retail bot on Kik

Image Credit: H&M

“In some cases, bot developers have had to roll back NLP or A.I. features, since it hurt the user experience rather than helping,” Sheth said. “When I’m ordering fast food, I don’t need to discuss the philosophy of life, I just want Combo #2. Depending on the context, less is more.”

Kosinski agreed brands need a well-defined bot strategy, whether it’s for customer service, ecommerce, or consumer engagement. From there, brands can adapt based on audience use and preferences.

Cohen-Aslatei agreed that brands must figure out what functionality customers most value. “Bots work best with limited functionality that is highly repeatable. It is not meant to replace an app or provide highly customized services, but to provide time efficiency and convenience to users,” he said.

Still, it’s early days for bots, and they will take time to develop. “I think there’s an expectation that every bot will be like Samantha in “Her,” but that’s not the case. It will take a lot of time for smart people to build A.I.,” said Paul Gray, director of platform services at chat network Kik. “I think it’s better to keep bots more focused and functional rather than to see everything. Tay was trying to be a realistic, lifelike entity, and obviously people were able to game it and make it kind of evil. The risk there is it is obviously a big project. If you’re trying to build an amazing bot, look at the bots on our platform. They are simple, focused, humble, and really beginning to get out there.”

And these, he said, include a “very straightforward” H&M bot that asks consumers what they are looking for and helps guide them.“It’s just going through a conversation. There’s nothing this bot could say or do that goes off the rails,” Gray said.

Dave Cotter, CEO of text-message-based retail service ReplyYes, also believes that brands must focus when it comes to bots. “Trying to be all things to all people means you’re nothing to most everyone,” Cotter said. “We think vertically focused retail has worked well because it’s a clearly defined product that has a commonly used set of words, and it also allows us to set the right expectations with our customers.”

4. Start with a strategy

And, as with any endeavor, brands shouldn’t simply put a bot out there just to have a bot out there. “You have to have a use case and a strategy to go along with it, but you have to balance being open to new platforms and opportunities. Finding balance is important,” Kosinski said. “Brands that know their customers and what they want…offer a unique opportunity for brands to become more than just a brand and become one of their friends.”

Similarly, Gray recommended marketers carefully consider what the problem or opportunity is that they are trying to address. “For marketers, that’s often ‘Is this something about generating awareness or managing a reputation or driving a direct action?’” Gray said. “That can help plan out what the bot should do. It’s not too dissimilar over different eras of technology, like in the ’90s when we were asking ‘What is a website for and what does it do?’ For brands, it’s about awareness and getting consumers to buy.”

There are many questions brands and marketers should ask themselves, said Francis of Alter Agents. They include: What purpose will this bot serve? Is this purpose a necessity? Does it provide value that customers are not currently getting from this brand? And how long do we plan to employ this bot?

5. Focus your bot on a single campaign

Similarly, Gray said it’s not about making a bot for a whole brand immediately — it’s about creating a bot for a single campaign. He pointed to Burger King, which he said has released bots related to specific products, like Chicken Fries. This takes away some of the pressure to build one amazing bot right out of the gate.

“And once the campaign is done, you can take a break from that bot,” Gray said. “It’s why movies do things with bots and chat. There’s a big lead-up to whatever it is, then it’s finished and drops off.”

6. Test and learn

This image shows a a screenshot of Microsoft's Tay bot.

Above: Screenshot of Microsoft’s Tay bot.

Image Credit: Screenshot

Pypestream’s Smullen noted that businesses seeking a bot to answer a broad spectrum of questions in inbound types of communication will need significant resources. These companies must be willing to invest time and money into programming the bot to handle all kinds of queries and permutations in accordance with how a human being would have responded.

For his part, Francis said currents bots should have a shelf life, since they need to be constantly updated.

And Rod McLeod, PR Guy at Kik, noted that brands need to realize bots are a new experience that requires experimentation. He points to social media: “We’ve seen some examples where people have used broadcast methods like it is a one-way channel, which is not what it’s meant to be. Brands need to take chances and experiment.”

It’s a reminder that brands still have much to learn about bots. “Although bot technology has been around for a while, its new use within social and messenger apps has not yet been fully proven,” Cohen-Aslatei said. “Slow bot response time, bot confusion, and limited product and service offerings have caused some consumer frustration,” he said. “In the future, brands should be able to access more robust data and use it to better tailor chatbots to customer preferences. For now, brands should certainly test chatbots, as long as the benefit of improved customer service outweighs any shortcomings of the technology.”

7. Don’t be afraid to partner

“Messaging is the wild, wild West,” said Collaboratory’s Kosinski. “It’s scary, but it’s an opportunity and a unique time for brands to partner with startups that have expertise and that have connections with Facebook and Messenger and understand A.I. It’s a great opportunity to reach out to startups and learn from them, partner with them and to grow in parallel.” Gray agreed, saying that after Microsoft and Facebook made their bot announcements, “People are expecting a great out-of-the-box experience, but they need to be patient and partner with [a player like] Kik and other developers, so bots get better.”

8. Interaction must be meaningful, not chatty

Bots have to enable meaningful interactions that help customers achieve their goals with the fewest distractions. “Bots must set the right expectation and deliver the promise,” Gupshup’s Sheth said. “Bots must avoid intrusive spam and excessive chattiness, otherwise you risk losing customer engagement.”

“Spam is a four-letter word for goodbye,” Cotter of ReplyYes said. “Respect the trust customers grant you by allowing you into their text or app messaging inbox.”

9. Pair bots with humans

The nuances of human conversation can be hard for bots to handle, and when bots can’t handle those nuances, it’s a terrible customer experience. “Bots are good at tasks, not conversations,” Cotter added. He notes that ReplyYes’ first so-called conversational commerce channel has human operators as well, “to jump in when bots aren’t right for the job.”

10. Don’t try to pass bots off as humans

Joaquin Phoenix in the AI love story "Her."

Above: Joaquin Phoenix in the AI love story “Her.”

Image Credit: Annapurna Pictures

Another pro tip: Consumers are smart, so don’t pretend your bots are people. “Humans have a sense when companies are trying to trick them, and it breeds mistrust,” Cotter warned.

11. Don’t forget geography

Finally, consider where your target audience is based. Platform usage varies by geography. What’s popular for messaging in North America may be different in Asia.

“Brands have to understand usage is very different based on geographic location,” Kosinski said. He estimated that there are probably eight or nine large platforms that are the leaders in messaging, but it’s a market that varies based on geography and user age. “Europe is dominated by Viber and WhatsApp, and Germany is dominated by Telegram, which has unique privacy features that Germans are really attracted to, whereas Line has about 96 percent in South Korea and WeChat is huge in China.”

Plan for success

While bots today are a far cry from Spike Jonze’s vision of Samantha in “Her,” mean app usage is likely to decline. Brands have little choice but to master yet another environment: bots.

Fortunately, bots are still nascent, and there’s plenty of opportunity for brands and marketers to figure out what works. Some efforts will fail, just as other efforts will succeed.

The brands that embrace bots first, work with knowledgeable partners, ensure their bot experiences are focused, and employ bots in conjunction with people — without trying to pass them off as people — are the ones best poised for success.