Chatbots: What They Are, and Where They’re Headed

On July 26, 2017, 30STF attended the Chatbot Growth and Engagement Roundtable in NYC, featuring 16 leaders of the AI and chatbot conversation. It was hosted by PluggedIn BD at the Cushman and Wakefield offices.

For the chatbot uninitiated, for those who have not attended the roundtable, the leading perception might be that chatbots are a gimmick that’s all show and no substance. Maybe you’ve interacted with a customer service chatbot and not even known it. But the Chatbot Growth and Engagement Roundtable revealed that the future of chatbots is frictionless communication. And that future is now.

When handled properly and in the right hands (these right hands being defined as savvy marketers who understand what their customers need), chatbots can provide a lot of value. They can simplify transactions, personalize sales, and expedite customer service (or at least eradicate horrible over the phone support). 

When done poorly, they can be unethical, unhelpful, or (perhaps worst of all) ignored. They can overstep boundaries, be mismanaged, or simply go out of style. They can be seen a just a gimmick, or a cool technological trick.

When done well, they can immensely improve relationships between customers and businesses, and make things frictionless. They’re more than a gimmick- they’re a necessary business decision and a huge win for customers.

Here is what we learned about chatbots from the Chatbot Growth and Engagement Roundtable.

Chatbots are helpful tools for companies and customers

According to nuclear scientist and AI specialist Riza Berkan, we currently live in a world of ‘Static Knowledge’. We have paper censuses, FAQs, customer service questionnaires. These tools for providing or obtaining information are not dynamic. They’re created, ingested, spit out, and left alone.

Chatbots can, first and foremost, bring us into a world of Interactive Knowledge. Businesses can learn more about their customers, and customers can learn to engage with smarter, more personalized, and more helpful service. 

To fully reach their potential, chatbots need to understand how humans interact- one of the great challenges of Turing Test computer science. They need to have an abstraction capability that can allow them to answer similar questions posed in different ways. They need to have innate logic about presenting information in the best medium (ie. it wouldn’t help to send the 10 page terms of service document when the customer asked a question about refunds). And they need to be frictionless.

Currently, most people likely get our customer service questions answered via FAQ (static and devoid of a lot of key information) or via a call-center based customer service agent (because everyone just loves hold music). If chatbots can bridge the ease of use of an FAQ with the helpfulness of a real human, chatbots can create a frictionless method of helping the customer.

And if they can do that, they can help businesses. It’s a three step process: Appreciation, accumulation, evaluation/execution.

The three steps to a company’s chatbot adoption


Businesses need to understand the enormous value in data that chatbots can generate. By creating bots that keep customers engaged, businesses can learn about their tastes and habits. By providing offers tailored to the customer, a business can grow sales. And by algorithm-ising (read: the creation of an algorithm from previously manual tasks) the customer service process, businesses can retain customers through excellent, frictionless, service.


Your bot it built, now it’s time to get data. As Stephen Nemeth, VP Director of Strategy and Innovation at Carat, put it, “it’s about the ability to understand people throughout the day and the buying journey”. Interactive consumer panels help get more iterative product reviews and builds. Understanding emotions can help improve and tailor marketing efforts. Imagine a world in which every customer interaction can be recorded, coded, and extrapolated upon. That can happen with chatbots in a way neither static websites nor human customer service agents could ever do.


It all comes down to getting things done. Once all of this information is collected, it’s the company’s job to make the process/product/service better for the customer. Better understanding the most common customer service questions asked can help tailor responses. Evaluating a customer’s comprehension of your product (novice vs. expert) can allow your chatbot to tailor its responses to be most helpful to the customer. And learning about your customer’s style and preference allows your chatbot to recommend (and ideally sell) very specific products.

So that’s what chatbots can do. They can immensely improve relationships between customers and businesses, and make things frictionless. But, Stephen Nemeth notes that this comes with a warning label.

“If your customer service sucks, chatbots are not going to improve it.”

Just like “nothing kills a bad product faster than good marketing” (Brad Brinegar, McKinney), it’s all about creating a good system to translate into a good chatbot, not the other way around.

Chatbots are not easily adopted by marketers

Marketers are usually the most experimental and creative adopters of new technologies. They are willing to experiment with 4 second Snapchat ads, even though it’s a million miles away from 60 second tv ads. They’ll do customized QR codesbranded VR experiences, and Foursquare checkins galore.

But many marketers are resistant to incorporating chatbots into their companies. Part of the reason is that no one really knows what to do with them. They can be overlooked as a party trick or an expensive gimmick.

Before launching on the expensive and complex route of launching a chatbot, marketers and teams need to understand what they’re doing. They should ask themselves: Why do I want to build a bot? What is it going to do? Where does it fit into my ecosystem?

Chatbots today are a lot like iPhone apps in 2008

There’s an obvious recent historical parallel in which technology presented itself to businesses in an exciting and new manner. Remember the excitement around the launch of the App Store? Yeah, so do many companies. And not always positively. 

When the iPhone App Store was released in 2008, every business in the 1st world universe (and many in the 2nd and 3rd worlds) believed that they needed to build their own app. In 2011, the New York Times released an article proclaiming news that the millionth app was launched, with over 15,000 new apps introduced every week. Just 4 years later, the International Business Times stated that that number had gone down to just 7,000 new apps every week. People and companies realized that apps aren’t a must have. Not every brand, store, market, company, tool needs (or can be serviced by) it’s own proprietary app. Technology that isn’t used becomes useless, and depreciates over time. Learning from early mistakes, companies may be more wary.

Additionally, the data to support chatbot value isn’t quite there yet. As Gabe Weiss, Chief Strategy Officer at RAPP put it, “98% of users are on Messenger or that the average person chats with 15 people a day doesn’t make a marketer any more willing to build a chatbot.” Just because a lot of people use Messenger as an app with other humans, it does not mean that they will be willing to communicate in the same way with an AI chat bot. Using Messenger use data to pitch the value of chatbots doesn’t make sense. There needs to be a more exciting pitch to marketers to the importance of chatbots.

Having successful chatbots is about good strategy

Part of good execution is good strategy. Follow the first rule of content marketing: don’t throw money at a place your customers are not. If your customers are not already first adopters or fast followers of technology, they’re not likely to pick up on chatbots until they’re more integrated. 

Even if a company were to develop a chatbot, and one with all the bells and whistles to be successful, chatbots are nothing more than the next opt-in media. They need the customer’s participation in order to work effectively. So companies can’t just build and put it out there. According to Anna Nicanorova, Director at Annalect Labs, chatbots, in contrast to apps, are about “doing a little work upfront, and lots of changing throughout. The value of chatbots comes from learning about how to improve.” 

And improvement comes from engagement. So how can a company drive engagement? Gabe Weiss’s company helped Air France with a brilliant marketing effort where folks who chatted with Mr. Miles (Europe’s chatbot version of the most interesting man in the world) became eligible for a 500,000 mile prize. This led to enormous engagement which in turn created huge amounts of information gathered, and many happy and impressed customers.

Which leads us to other examples of chatbots done well.

Services must mix digital and physical
The best service: a mix of digital and physical

Who is doing chatbots well?

As with any product, the best version is the one that solves a major pain point. Whether for a company or a customer, here are some examples of well made chatbots.

For the company:

Redbull- training and on boarding new employees. Daniel Ilkovich, CEO of chatbot shop Dexter, told of how Redbull is using chatbots to train its internal employees. Creating an easily navigable and frictionless way for young employees to ask questions and get answers saves enormous amounts of money and human capital in training. builds chatbots for brand mascots and movie characters. Imagine a 5 year old boy being able to chat with the real Spiderman. Or a physics student having a chat with Einstein. Imperson creates branded chatbot robots to help companies promote their brands and initiatives.

For the customer:

Gerber- a human guided chatbot named Dorothy. This is a very on-brand and human supported chatbot to be our “Personal Baby Expert.” Dorothy helps customers learn about nutrition, lactation, and sleep. Plus, it’s real humans in the background which gives customers a great experience while allowing the bot AI to learn from standard answers given by humans. Eventually, with enough inputs, Dorothy can become smart enough to hold a conversation without humans watching her back.

Claire- making travel booking easier for SMB. Claire is an AI assistant that helps small and medium – sized businesses manage, control and automate their corporate travel. Using any messenger platform, Claire books business trips within seconds. She knows the team’s travel policy as well as their preferences. Claire provides personal assistance 24/7 and solves any problem during the journey. On an organizational level, Claire provides real-time insights into travel analytics and helps management to reduce travel expenses systematically. So through a know-it-all travel chatbot, Claire can save employees time, business money, and tracks more industry data than any human HR team can.

Chatbots can be hyper personal

One of the great values of chatbots is, in fact, their ability to be personalized. It’s something we mentioned at the very start of the article, and it’s one of the reasons companies have invested billions in their creation.  

Just like human speech varies according to who you’re speaking with (your best friend or your grandma) and the context (the New York ballet or an amusement park), chatbots should be able to do the same. They should be able to identify our mood and needs based on our questions and text, and respond accordingly. Curation is how they can provide the most value to the customer, and how they can appear to be most genuine, enjoyable, and frictionless conversationalists. So good, that they may almost seem human…

Though, as said by Judy Shapiro, CMO of Chattify:

“The objective of the chatbot should never be to be human. It’s to provide a service” 

Personalization leads to segmentation and bias

So we don’t need a chatbot passing a Turing Test, we need a bot helping people. And personalization is a big part of that. One of the greatest advances of the chatbot age is a massive change in segmentation. While before companies attempted to group users based on demographics (race, age, gender), chatbots are able to take the double blind approach and identify users solely by their needs, based on the questions they ask. So instead of a customer being a Chinese woman, 35 years old, a chatbot can identify a customer as a 1st time user of this product with above average technical knowledge and large budget. This clarification is more useful to a company than the former could ever be.

Replacing demographics with need-based curation is a huge accomplishment, but a huge task. It’s easy to divide customers by what they look like. It’s hard to differentiate them by true need. It requires truly neutral understanding on behalf of the robot, a job not so easily accomplished.

The internet has actively discussed inherent bias in human-developed technologies. Here is a CNN Money video to brush you up on the biases of tech.

The premise is that human developed initiatives have the biases of their creator. Men build apps that are more tailored to men than women. Black women build businesses that white men have trouble understanding. And Frankenstein’s monster had the sweet, misunderstood personality of his creator.

AI only expedites developer biases

And with AI, when technology builds upon itself and continues to improve exponentially, biases accumulate and begin to get exceedingly impactful. One example mentioned multiple times at the roundtable came from the recent US election. According to chatbots developed mostly in New York and California (highly liberal states), America was voting blue, and Mrs. Clinton had the presidency in the bag. But the innate bias of the chatbots made them unable to interpret answers in a neutral enough way to appreciate subtleties that would make a voter choose Mr. Trump. And so when election day came along and shocked many chatbot creators, all it took was a look inside the code to spot the biases that mistakenly evaluated the accumulated answers.

So biases exist, but they aren’t necessarily bad. They’re part of our world. “Every brand is a bias,” said Stephen Nemeth. “If the world is neutral, the world is not interesting.” Branded bots need bias- imagine a neutral chatbot trying to sell a woman makeup, without any emotion or opinion. That would be a poor experience for the female customer, to say the least. Products need bias- would anyone use the Fox News chatbot if it existed without a conservative tint? Or interact with a Comedy Central chatbot who didn’t have a sense of humor? The key is to understand the bias innate in your product so that you can serve your customers in the best manner. 

What is the future of chatbots?

Plenty of 21st century technologies have were promised to change our world are now defunct. Will chatbots go the way of the QR code, or the Avatar? A cool technology that was never really able to scale properly?

The overwhelming answer of this (admittedly biased) roundtable was no. Learning from the mad times of iOS app development is helpful in making good decisions. Being integrated with huge apps like iMessage and Messenger limits the self-destructive capabilities of chatbots. And the market is so big that chatbots will find their place.

A lot comes from our changing tastes and needs. As generations get more tech-friendly, the expectations will come down. We’ll become more comfortable with machines that aren’t exactly human, and aren’t entirely robot. We’ll be okay sharing more with a machine than we ever did with a human customer service rep, even some of our most private information, like medical records and financial stats.

The future of chatbots isn’t written in stone. If mismanaged, it can end in an anticlimactic disappointment. On the other hand, it can change how we interact with every business in the world. Overall, the industry is young and expectations for its future are no more certain than a talented 8th grade pitcher’s career prospects. Overall, chatbots have a lot of value to add to the world today, and a lot of problems to solve in our future.