Vladimir Bukovsky, who died at 76 last month in Cambridge, England, leaves a legacy and example for anyone who feels they can’t go along with the party line. As a young man in Moscow, he persisted in publishing and reading forbidden works in the face of certain incarceration. He racked up 12 years in labor camps, prisons and phony psychiatric wards before an exasperated Kremlin traded him to the West in a prisoner exchange in 1976.

Bukovsky’s final book ​Judgment in Moscow​, belatedly published in English a few months before his death, lays out a paper trail of Kremlin documents he stole in 1992. They illustrate both Moscow’s covert reach around the world and its leaders’ petulant intolerance to criticism. After reading their meeting minutes, it’s too bad the Central Committee wasn’t on Twitter to keep us entertained.

Judgment, w​hose English edition I prepared for publication as a volunteer after 24 years of it being dodged by wary publishing houses, is built around thousands of pages of top secret Communist Party of the Soviet Union documents that the ex-pat Bukovsky scanned when then-President Yeltsin brought the exiled writer back to Moscow to assist in a court trial brought by deposed Communist Party officers.

Poring through decades of classified papers in a guarded archive, Bukovsky used a newly developed Japanese scanner — unrecognized by those watching him — to make copies of classified records. His takeaway ranges from an early scribble by Stalin, glibly raising the quota for executions, to Gorbachev’s ​fin de siecle​ response when told in a meeting about the Tiananmen Square massacre of three thousand unarmed protesters: “Three thousand … so what.”

In between are decades of recorded meetings and communications between Soviet officials and leaders in Western Europe. At best, the USSR’s neighbors are trying to keep peace with the superpower next door. But too often they’re secretly courting Moscow’s mafia in hopes of a seat at the table.

The book’s record of secret dealing goes back to 1969, when West German minister Egon Bahr seeks “a series of unofficial negotiations” out of sight from his official negotiators in Bonn. His goal is to keep political opponents from having anything to use against him. Two decades later, French President Mitterand proposes to the struggling Gorbachev “a new Union” that would prevent the collapse of the USSR through an alliance between Paris and Moscow. More comically, the head of Sweden’s leading socialist party is so covertly helpful to the Central Committee’s agenda that in one meeting, they agree to send him a birthday present.

All along, the Kremlin approves covert funding and aid to Western activist movements that keep their opponents off balance. They ship Russian-made paper to South American propaganda publishers. They quietly aid advocates of nuclear disarmament in the UK — you go first, England! They propose to fund and advise the Black Panthers in the U.S., in a formal request that begins by invoking capitalist oppression of African-Americans in America but devolves into what a great idea this is to keep Nixon’s attention off foreign policy.

Judgment​, which Bukovsky wrote in his native Russian in 1995, was translated into several other languages in the mid-90’s. The French translation sold well. But U.S. and UK publishers balked at putting their imprint on an English translation. British publishers were bluntly threatened under that country’s tort-friendly libel laws of the time.

In America, ​Judgment​ faced a different roadblock. The new edition includes correspondence from a long back-and-forth with a senior editor at Random House. He wants Bukovsky to rewrite the book to be less accusatory of the few Americans named in the documents — Secretary of State Cyrus Vance, filmmaker Francis Ford Coppola — whom Bukovsky felt shouldn’t have negotiated with the Communists at all, whether on terms of disarmament or to propose a documentary on it. The editor seems less worried about lawsuits than about being criticized for an insufficiently left-leaning book.

A quarter century later, in an era of hectoring TV hosts and Internet mobs, that fear seems both antiquated and evergreen. But ​Judgment’​ s central theme from 1995 is no longer controversial: If we don’t prosecute Soviet crimes against humanity as we did the defeated Nazis (the book’s title is a reference to the film ​Judgment at Nuremberg)​ , the abuses and subterfuge will never end. The same criminal gangs will rebrand and return to power, Bukovsky warned after seeing old nemeses from the 70’s still in office in Moscow. Their heirs, he warned, will continue to use people as raw materials, grinding them underfoot and extracting the joy from their lives.

Bukovsky didn’t get his trial, but he did leave a dual legacy. First, ​Judgment​ is a lengthy tour behind the curtains at the Kremlin. It illuminates the behavior of today’s occupants, many of whom are the KGB-era veterans including Vladimir Putin.

If you’re looking for a clockwork game of chess being orchestrated from Moscow in today’s news out of D.C., ​Judgment ​reveals Russian influence to be more like a mean game of darts. They fund one incompetent venture after another. They occupy Afghanistan and believe reports that everything’s under control there. They hang onto separatist Poland but also carry tens of billions of dollars in Polish debt, even as President Reagan is driving them into military-budget bankruptcy with his dubious Strategic Defense Initiative. Notably absent is any sense of crisis around Chernobyl, which Gorbachev himself now claims scuttled the regime’s remaining credibility with its citizens. Even Gorbachev proves a different man behind closed doors. His take on the gunning down of protesters in Soviet Georgia: “You have to put people in their place.”

As Bukovsky’s editor, I asked him to write an epilogue for ​Judgment​, tying his early-90’s book to current events. “If you read the book,” he dismissed me, “you don’t need me to explain anything.” True, the Soviet ​modus operandi​ — scattershot and self-serving — can be spotted in modern Russian attempts to sway American politics. They’re not grooming Manchurian candidates, they’re simply boosting disruptors. Their operative who sought a deal with Trump’s campaign wasn’t looking to get the U.S. out of Syria, or help Russia’s enormous oil and gas industry by scaling back American fracking. No, Natalia Veselnitskaya’s target was the Magnitsky Act, which authorizes the Senate to freeze billions of dollars in personal assets of Russian officials in response to human rights abuses. Magnitsky is the closest thing to a trial and sentence that Bukovsky lived to see.

The Putin regime’s focus on their personal bank accounts leads to the more universal legacy that Vladimir Bukovsky leaves us: As a man who spent his twenties and thirties repeatedly incarcerated for his words (and for smuggling phony diagnoses of “schizophrenic” dissidents into the hands of doctors worldwide), he had a short fuse for would-be dissidents whose utmost priority was their own comfort.

As a young man, he had refused the chance to avoid prison via a public apology. He instead gave a courtroom oration that novelist Vladimir Nabokov called “a ​heroic speech in defense of freedom” before being led away to labor camp. In​ Judgment,​ he berates Americans as soft, capitulating followers who backpedal from their convictions—not to avoid labor camp, but to keep their comfy careers. He compares us to the professional ​intelligentsia​ of the Soviet era, who strode safely in and out of the Institute of Literature past unbending poet and playwright Andrei Platonov, whose terms of censure by the government let him work at the Institute, but as a watchman in a hut outside:

All these people were phony “authorities” whose works did not, and could not, survive the regime. Listening now to the keening and wailing of the intelligentsia about how they suffered, how they were forced to lie, I am deeply puzzled: Why was it necessary to become writers, professors or academics at any price? As we can see here, talent has nothing to do with it; talent survives being a watchman. Everyone has a choice. But no, nobody was prepared to become a watchman, everyone wanted to suffer in comfort. Everyone needed an honorable justification for their own conformism.

By contrast, Bukovsky wondered why Hollywood’s actual communists of the 1950’s didn’t stand up to Joe McCarthy. “They faced no threat of camps, or torture, or destruction. At worst a loss of their jobs. It’s curious how the majority of them broke so shamefully, pointing fingers at their friends and neighbors and lying under oath. Only a few refused to speak. Suffering heroes indeed!”

In his writings, Bukovsky was forgiving to fellow dissidents who signed phony recantations and diagnoses of mental illness to get out of lifetime lockup. He was touchingly sympathetic to his prison guards, whom he saw dehumanized and made miserable by years of bullying and torturing fellow humans under orders. He saved his perpetual spite for the sellouts who lie about their beliefs as a career move. The skilled naysayers who always know where the approved bounds of dissent are drawn. The poseurs who curate their public personas to appear defiant while posing no real threat to either the state or the mob.

Many won’t agree with Bukovsky’s take on world events, but reading him conveys an inarguable insight: Whatever you believe in, you’ll know when you’re up against the wall to deny it in word or in deed. At that moment, he wrote, change is wrought by people like himself who are less brave than stubborn: “With his back to the wall, a man cannot sacrifice a part of himself, cannot split himself up. And an astonishing thing happens. In fighting to preserve his integrity, he is simultaneously fighting for his people. It is such individuals who win the right for their communities to live — even, perhaps, if they are not thinking of it at the time.”

Paul Boutin is Executive Editor at Ninth of November Press, which published ​Judgment in Moscow​ in English.

In his annual essay to this year’s graduating class, Microsoft founder and technology visionary Bill Gates — I hate Windows as much as you, but the guy knows an opportunity when he sees one— listed artificial intelligence as the first career he would recommend to new college grads. “We have only begun to tap into all the ways it will make people’s lives more productive and creative,” Gates wrote.

At the same time, you’ve surely noticed that “AI” has become this year’s buzzword of choice for marketers and second-rate news writers. It won’t last, but Gates is right: behind the hype, artificial intelligence, hard as it is to define, has been gradually becoming more prevalent in our lives. “Machine intelligence is everywhere,” MIT Technology Review chief Jason Pontin told Chatbots Magazine in April, “and we couldn’t live without it now.”

Couldn’t live without it? He’s not exaggerating. Our food still grows out of the ground in ways we don’t understand, but there’s a lot of mind-boggling math involved in getting it to the table, fresh and affordable — everything from weather forecasts to market forecasts, buying and selling, even routing the trucks that deliver it. AI may soon do the driving, too. If I were a trucker in 2017 as opposed to the free-wheeling Seventies, I might be relieved to hear that the Qualcomm box on my dashboard is going to stop telling me how to do my job all day and take the wheel itself. Let me know when we’re there, I’ve got some Facebook to catch up on.

But note the way AI is usually presented: It’s going to take away our jobs. Will AI create jobs for anyone but machine learning engineers and data scientists?

Yes! Artificial intelligence, like its best buddy robotics, will indeed eliminate a lot of roles still held by humans. But it also will create or enhance roles that only a human can fill, jobs that aren’t computer programming or maintenance. Here are the two most obvious types of work that won’t be automated soon.

The “Hard” Skills: Training and Analysis

An algorithm that makes decisions for people is only as smart as the people who tell it what to do. Current AI requires a lot of training on a lot of data. You need to know what the computer should choose, and what counts as a wise decision. To do that, you need to understand the territory, whether that’s medicine, logistics, investing, or getting people to bond with a chatbot.

What information goes into a decision? What level of certainty do you need before you act? Where should the lines be drawn between letting the machine do everything, and bringing in a live human to take over? Getting these right will be crucial to any AI program’s success. AI can’t train itself.

And for many AI applications, the whole point is to present information to a human for interpretation and analysis. Just as a human needs to define success to the computer, a human needs to figure out if the results are, indeed, success. And humans can spot patterns and anomalies in reports that turn out to be important, patterns which the computer was not trained to look for.

So go ahead and let a computer take your job. Figure out how to promote yourself to its manager.

The “Soft” Skills: Human Language, Human Emotions

Have you noticed that Chatbots Magazine isn’t written by the bots themselves? Chatbots aren’t even good at chat. People have the misconception that computer programming languages are complex and hard.

On the contrary, Java and Python are simple and rigid compared to English or Chinese. I loved hearing the story of two over-age-50 paper mill workers in Maine who were told it would take months for a factory robot’s programmers to fly out from Scandinavia to reprogram it to stop labeling their proudly finished products at an amateurishly crooked angle. They sat down with the manual on their own initiative and fixed the bug. Making great paper is much harder than learning to code.

Human languages, by contrast, are complex, nuanced, flexible, and malleable, and people are much more clever than computers at using and understanding them to communicate ideas. I can verb a noun, and see? You understood what I just said. Try that on Google Translate.

Any career that involves a lot of spoken and/or written words is one that won’t be automated anytime soon. (You do have to worry about economic upheavals wrought by technology. Ask a writer. 🙂) The same goes for roles that require empathy with live human emotions, not mere sentiment analysis of a text message or image. What’s the difference between Woebot and a trained, experienced human therapist? The therapist can tell when you’re lying.

Which Careers Will Last?

Here’s a simple guide if you’re looking to map out a career in an AI-infused world: If it gets boring fast, let the machines do it. If it involves interacting with emotional, complicated, self-contradictory, dishonest human beings and somehow everyone has to come out feeling they’ve won, you’ve got a job.

If that job involves fearlessly working with AI you can barely understand, you’ve got a career.

Click the 💚 below to tell Medium to suggest this story to others looking to read about careers in the AI future. See? Medium’s computers need your help!

AI Careers For Dummies was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

If you’ve been rolling your eyes waiting for a real superstar to adopt a chatbot in earnest, click here. Katy Perry’s Messenger bot, developed by the boutique firm that built them for Snoop Dogg and the fictionally naughty Christian Grey from Fifty Shades of Grey, is now live where some 66 million followers can find it. Perry’s publicity machine. in high gear to promote her new album Witness released last week, hasn’t pitched the media about the bot, because the bot is but a tool to promote the album and its tour. You’ll notice that Katybot is intent on getting your location for show info.

Persona co-founder and bot designer Josh Bocanegra, who takes pride in focusing on little personal details that make a bot seem more human — Christina Milian’s bot pretends to be typing for an instant before it replies, rather than robo-responding instantaneously — isn’t ready to talk about Perry’s bot in detail yet. But you can see the attention to detail in the way it explains clearly to new followers how to interact with its guided conversations.

“Sorry, we’re not disclosing anything at the moment,” Bocanegra wrote in response to an email we sent with a few burning questions. “However, I would say our goals and choices are data driven, not opinion-based.”

The Year of Early Adoption Is Over

Our choices are data-driven. Josh (calling him “Bocanegra” sounds like we’re frenemies) couldn’t have said that a year ago, because there was no data. The rush of celebrities, big brands, and nobodies who flooded Messenger with thousands of beginner-bots a year ago petered out when marketers with goals didn’t see those goals being met. Internet adventurers lost interest when the first built-int0-Facebook chatbots didn’t live up to the blatant hype that they would engage in friendly small talk with us while selling us shoes chosen personally by artificial intelligence.

Those weren’t data-driven plans based on a chart from Forrester Research. They were the audacity of Hope. But a year later, we’ve got numbers to work with. Chatbots have a measured 15 to 60 percent response rate, numbers that mailing list managers have trouble believing. Some brands claim to get as many responses from their bot as they do from their Facebook page. No names here, but I checked a few big bot brands and many aren’t bothering to run Google ads. The bot is where it’s at.

When Facebook used Maroon 5 — a client of Octane AI who publish this site — as a go-to example of chatbot marketing at the F8 developer conference in April, you could feel the measured response. Sure, M5 had used the bot to launch another hit single, and Adam Levine has a lovely falsetto, but where are the supergiant power players? Where’s Lady Gaga’s bot? Taylor Swift’s? Their Facebook pages can be summed up as a Shop Now button.

But now Katy Perry has a bot. Katy Perry isn’t sort-of big, nor is anyone wondering if she has staying power. It’s been almost a decade since anyone believed “I Kissed a Girl” would be her first and last hit. She’s racked up nine #1 singles. Witness, released last Thursday, is likely to debut at #1 on Billboard’s album chart this week, even though its singles haven’t blown up and Spin hates it.

The release of Witness is the most important, carefully orchestrated marketing event in Hollywood this month. Did you notice that Taylor Swift finally let Spotify carry her music beginning the night of Witness’ release? The album’s marketing armada includes a carefully planned bot built by a team that’s spent the past year establishing itself as the new genre’s hitmakers.

I think we’re out of the early adopter phase.

Click the 💚 below to tell Medium to recommend this story to people looking for stories about chatbots, Katy, and hey Taylor Swift is on Spotify now.

Katy Perry Removes All Doubt About Chatbots Going Mainstream was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Would you have your paychecks deposited to Google, Apple or Facebook instead of Wells Fargo or Bank of America? Your kids might. The single outstanding factor customer seek in a banking brand is trust with their money.

Chatbots Magazine has run many articles by people who build and deploy bots for big banks, or for banking startups. These seven offer a quick way to get up to speed.

“Taking the monotone and repetitive tasks out.”

Banking with Artificial Intelligence

“Let your customers use the words they are familiar with and use every day, not the quasi-words you tell them to use.”

Banks Have Been Deploying Chatbots for a Decade and Nobody Talks About It

“We can fit these conversations into our lives, rather than having them interrupt us.”

Conversational Banking: From Branches to Bots

“Forty percent said they would switch to Google, Apple or Facebook if they were a bank.”

Would You Quit Your Bank for Google? Q&A with Kasisto’s Zor Gorelov

“Banks are often thinking: The more features we have, the better.”

What We Learned Designing a Chatbot for Banking

“Much like we now use electricity or water to solve problems, artificial intelligence and computing will be a utility.”

Artificial Intelligence(AI) and FinTech.

“Conversational software will eliminate the cognitive load required to plan and manage finances.”

Bot thoughts from the world’s largest fintech conference

“Expect some strange responses.”

Five Insights from Releasing a Banking Chatbot in Brooklyn

Click the 💚 below to tell Medium to recommend this list to other readers interested in chatbots and banking.

The 7 Best Stories About Bots and Banking was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

A peer-reviewed study of 70 undergrad and grad students at Stanford has confirmed what the ELIZA chatbot seemed to demonstrate all the way back in 1964: Chatting with a bot that mimics basic therapy techniques can actually elevate the mood of a depressed Messenger user.

Woebot, the creation of Alison Darcy, an adjunct instructor at the Stanford School of Medicine, uses cognitive behavioral therapy (CBT) techniques to help people identify and manage symptoms of anxiety and depression, through identifying and changing patterns of distorted negative thinking — for example,“No one likes me.” Darcy is also CEO of Woebot Labs, the company that designed and built the bot.

No one is suggesting a bot is a replacement for a skilled human therapist — even Woebot makes that clear. But a chatbot could be a complementary tool: available 24/7 for an anonymous chat, wherever its client is at the moment, with a guarantee of privacy from its makers. No appointment necessary. And people have a natural tendency to treat chatbots as human, even when they know they’re not. As one writer noted, Woebot has an innate appeal to some people who might balk at a traditional therapy session.

The study found that students who talked to Woebot an average of 12 times over two weeks demonstrated reduced symptoms of anxiety and depression over those who were given a self-help ebook, which didn’t seem to help at all.

Moreover, Darcy believes that Woebot can reach people who wouldn’t get to a therapist at all. She has given a lot of press interviews this week, in which she’s laid out the idea behind Woebot, and the context in which it operates, in detail:

Why a chatbot?

“I tend to not think of this as a better way to do therapy. I think of this as an alternative option. What about the people who aren’t ready to talk to another person?”

“It’s almost borderline illegal to say this in my profession, but there’s a lot of noise in human relationships … the fear of being judged.”

“Woebot has the advantage of being not human; people won’t be afraid of burdening him at any time, they know he doesn’t sleep,” Darcy said. “I think they feel OK about sharing things because there’s nobody really behind it and they can be a bit more candid and open, and it removes that fear in the moment of sharing, which could be a lot of noise that distracts what you’re trying to say or think about.”

“It’s actually better to learn in small chunks on a daily basis than it is in one hour-long chunk in one specific place and time.”

What can Woebot do for people?

“One of the biggest tips I think for dealing with stress and anxiety is realize and understand how that anxiety is helping you first and foremost — because the objective is not to eliminate any emotion. You can’t selectively eliminate, unfortunately. It’s to dial it down to a healthy level. Understanding what’s adaptive about your emotion is the key.”

What’s the idea behind Woebot’s CBT techniques?

“Self-talk is a part of being human. But the kinds of thoughts that we have actually map onto the kinds of emotions we’re feeling.”

“The first thing that I do as a clinician is to help people see why they should be proud in a way. Either the emotion is serving a purpose or it says something really valuable about them as people. Sometimes, people feel depressed and it’s sort of an integrity thing; they’re depressed with the state of the world or something. Well, doesn’t that say something beautiful about you as a sensitive human being that is socially conscious? Understanding what’s adaptive about your emotion is the key, I think, to then have the ability to turn it down to a healthy level.”

How does Woebot actually work as a therapist?

“It really remembers what you say — things you might want to achieve, what you think your strengths might be. It helps you think through things like that and will bring them up at relevant times.”

“He’ll talk somebody through that process, and help them reframe those thoughts and really stick with them until they’ve come to the other end of that process, and that normally takes a full session a lot of the time; this is quite quick, I would say.”

“Some of the people were saying how much they had noticed their thinking had changed,” she said. “Those kind of things you really hope for as a clinician, but it’s really rare to have a response that fast. We’re just really excited about the potential.”

“If somebody’s mood is really flat and their energy’s low over time, [Woebot] will actually say, ‘I don’t know if I’m being particularly helpful. You might need some additional help. Let’s think about what that might look like,’”

Conclusions

From the the study, published in JMIR Mental Health’s June 6 issue:

While results should be viewed with some caution and the findings need to be replicated, this study nonetheless demonstrates that a text-based conversational agent designed to mirror therapeutic process has the potential to offer an alternative and engaging method of delivering CBT for some 10 million college students in the United States who experience debilitating anxiety and depression.

Disclaimers

Subjects’ moods were only tracked for two weeks, while clinical depression can be an ongoing ordeal that defies treatment and returns. Woebot was A/B tested against an ebook, not against a certified human therapist. Darcy is both a c0-author of the study and CEO of Woebot Labs. And while Woebot pledges that no humans will read anyone’s sessions, users’ messages pass through a lot of Internet points, including Facebook. A bot that meets Federal HIPAA standards for patient data privacy would probably require a standalone app.

Still, the study’s results are encouraging — for anyone who’s lived with depression, a bot is certainly worth a try versus another self-help book.

For further reading, see PC Magazine’s interview with Darcy.

Click the 💚 below to tell Medium to recommend this story to other readers interested in mental health technology soutions.

A Therapist Bot Actually Works? was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Selling online is one of the new economy’s hotter new career opportunities, but it requires a lot of shoveling that isn’t the fun part of selling: Scheduling ads, sending emails, posting to social media, writing the same promotional texts over and over for different ad networks with different user interfaces. Isn’t there a bot for that?

There is, at least for Shopify store managers. Shopify acquired bot startup Kit last year, and on Wednesday announced that Kit, formerly $10 per month, is now free. “We wanted to create a fully-automated persona that would take all the heavy lifting off their shoulders,” Kit founder Michael Perry said of online store managers.

Kit is basically like having a free intern who handles turning store inventory and sales stats into ads, social media posts, and emails about products. It’ll even send thank-you emails to customers. What’s important is that Kit works “end-to-end,” as Perry puts it — the store owner doesn’t have to do anything for the tasks to be begun or completed.

A quick checklist: Kit will update a store’s Facebook page. It will create either single-product or multi-product listings. It will retarget ads to store visitors through Facebook. It will post photo and video ads on Instagram. It will email customer lists about best-sellers or specials. It has templates for promoting new products, best-sellers, weak-sellers, or seasonal items. Its text (“copy” in ad-speak) are written by ad industry copywriters who know what sells, but can be customized and tested to improve performance. It sends thank-you emails to buyers. (Perry says they tried also scheduling Twitter ads, but found them not to be cost-effective.)

Kit is free — there’s no trial period or upgrade to a premium package to get essential functions. Store owners can, however, pay to use third-party apps (Kit calls them Skills) that offer search engine optimization, advanced retargeting, and other functions, but its role is basically to make selling on Shopify as automated as possible.

Kit follows one of the unwritten rules of bot-dom: A successful bot needs to complete tasks for humans, not just guide them through the steps. And a really good bot, once trained or configured, won’t need to be told what to do in the first place.

Click the 💚 below to tell Medium to share this story with readers looking to learn about bots and online selling.

Selling Online? Get a Free Intern was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Forrester report lists three big mistakes

Globally, 57 percent of companies either use chatbots already, or plan to do so in the coming year, according to a report by Forrester Research. “Many fail to use them properly due to common mistakes,” the report cautions. Those mistakes aren’t gaffes in content. They’re strategic blunders that need to be addressed from the top down.

Not clearly defining the bot’s purpose

Why does your organization have chatbot in the first place? As with smartphone apps, some organizations haven’t thought through their goals yet. It’s hard to have success when you’re not sure what success is.

Chatbots Magazine contributors, who’ve succeeded and failed in the real world, constantly stress the need to plan to have a plan before you build and launch a chatbot. That begins with making it clear to everyone involved why you are building and launching an automated representative of your brand.

Setting goals that are too ambitious for existing tech

Facebook’s launch of Bots for Messenger in April 2016 has become the event that chatbot followers roll their eyes at. It wasn’t only Facebook, it was everyone: Bots would carry on intelligent conversations, guide visitors to actions including purchases, and learn things about customers that even the customer didn’t know. Problematic human employees, begone!

This year has brought the backlash. It’s cool to ridicule chatbots as failures: Facebook’s M project can only respond to 30 percent of customer requests! Never mind that automating 30 percent of incoming requests is a customer support manager’s dream. Many companies expected bots to take over most or all need for human interaction. That’s another way to fail: Define success as something unattainable, rather than totally doable and valuable.

Launching before they are ready

Everyone in the world of Internet startups loves to go live. Ready, fire, aim! It can be invaluable for debugging and feedback that wouldn’t have happened on a test server. But your live testing should happen before the big launch, on a limited set of customers who might be forewarned that they are participating in a test.

Octane AI, which publishes Chatbots Magazine, encourages bot adopters to put up a live bot quickly, but the idea — especially for critical missions for large customer bases — is to experiment and develop before making a big marketing push or routing all customers to the bot as first contact.

Also be sure that you have the people and processes in place to deal with customers’ behavior in response to the bot’s launch. You don’t want to be overwhelmed by your own success.

Don’t Let It Stop You

Despite the report’s warning, Forrester concludes, “Marketers should not give up and stop experimenting, though: when planned and executed well (see data spotlight below), there are ways for chatbots to deliver business and customer value.”

In order to help companies both avoid excessive expectations, and plan now for future opportunities, Forrester created this infographic of what you can expect from chatbots, and when.

Two to three years before bots provide personalized shopping? It’s not the answer anyone wanted. But it’s quite possibly permission to stop, take a deep breath, and begin forming that planning team that will meet future chatbot tech with its game face — and a game plan.

Click the 💚 below to tell Medium to recommend this story to others interested in succeeding with chatbots.

Why Most Chatbots Fail was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

If you’re a marketer, chatbots can be an effective new channel for connecting with customers. But every single chatbot expert — the people who build bots with big followings — will tell you that it’s critical to treat them as a unique form of communication, rather than just routing your Twitter posts through your bot, too.

The closest marketing channel analogy for chatbots isn’t Twitter, or Facebook pages, or apps. It’s the trusty old mailing list, which tech pundits neglect but marketers rely on. Adapting your mailing list mojo to a bot requires you to understand how the two media channels are the same, and how they’re different.

Let’s walk through the list. We’ll stick to Messenger bots on Facebook for simplicity, since Messenger has more regular users than Twitter, Instagram and Snapchat combined, and because Chatbots Magazine has great connections to Messenger bot masters through our publisher, Octane AI.

How Chatbots Are Like Mailing Lists

People sign up to be contacted

Facebook bots are only allowed to contact members who’ve proactively clicked Follow to stay in touch, or who have posted to their page. They can terminate the connection anytime without even having to go to an unsubscribe page. So you can trust that recipients are interested in what you have to say.

People don’t consider being messaged a distraction

Did your team deploy a smartphone app? Did a lot of people who downloaded it turn off notifications? It’s not you — it’s the perception that an app is an impersonal piece of software, a machine.

By contrast, a chatbot’s direct message is perceived as a communication from another human being, even though recipients know it’s a bot. It’s like a mechanical friend, if you do it right.

People respond to CTAs — calls to action

You might have trouble believing the click-through rates (CTR) for Messenger bots: 15 to 60 percent is the official Octane stat. Compare that to mailing lists, most of which work hard to get a five percent CTR, and whose managers throw their arms up in victory at ten percent. Even premiere newspaper publishers would consider a 20 percent CTR a sign that their reporting system might be broken.

But right now, chatbot users are quick to tap on a link messaged to their phones. In fact, most responses occur within the first ten minutes of sending.

This is why so many marketers haven’t abandoned last year’s bots. There was a rush of irrational exuberance for chatbots in 2016, especially on Messenger. The bots still aren’t as AI-smart as many made them out to be. But there are brands and organizations who get as many responses from their chatbot audience as they do from their ten-times larger Facebook audience. And many use chatbots to court Millennials and younger customers, heavily-connected consumers to whom email and website ads are for old folks.

Be aware: The disdain for email and Web ads among the young is partly thanks to misuse of them by advertisers. The Interactive Advertising Bureau’s official assessment in 2015 was, “We messed up.” The marketing risk for chatbots isn’t that their natural language processing won’t get better, it’s that short-sighted campaigns will chase everyone away again.

Finding the right amount and frequency of content to send is a craft

Every audience has a sweet spot, or a zone, within which they’ll enjoy being contacted. Too few, and they may have lost interest by the time you show up on their phones again. Too many, and you’ll be dispatched with one thumb.

How Chatbots Are Not Like Mailing Lists, and This is Important

No one will complain that they didn’t sign up

Facebook, at least, is ruthless about its rules for whom a bot can message, when, and how often. One reason mailing list CTRs are so low — one in twenty-five people clicking a link — is that many mailing lists are padded with recipients who didn’t realize they signed up while making an online purchase, or whose names were bought or rented. The list provider may be stuffing the list with addresses they collected from the Net without asking, or bought from another vendor. Look at your inbox for proof: “You are receiving this message because you signed up to receive …” No, you didn’t.

By contrast, Facebook blocks bots from messaging members who have not messaged them first, or at least posted or commented on the bot’s associated page. This is one reason chatbot CTRs are in another dimension from mailing lists.

Most people read your messages immediately

Within ten minutes! Imagine a mailing list where half your recipients clicked the link you sent them as soon as it hit their inbox. That changes how you can and should use the chatbot channel. It’s great for getting instant feedback, or drawing an audience to a live event.

But you also have to consider whether your message will make sense as one that demands instant attention. Do you have a friend who sends you constant political posts while you’re trying to work? Do you quietly despise them for it? It’s like that.

Clockwork messaging looks like spam

Unless you’re a news publisher, sending everyone a message at the same time every morning is a good way to lose your audience. It’s like your friend who sends you a link every morning at about the same time, because they’re procrastinating at the office again. You agreed to connect with them, but expected real conversations. Your nagging annoyance at their predictable morning here’s another thing about Trump is exactly the feeling you don’t want to inspire.

But spontaneous updates are welcome

The Wall Street Journal chatbot’s daily news digests sometimes irritate me, because they’re not something I need to read right this second. I want them, but I flinch when they pop up on my screen during a business call. They would be much more welcome as an email.

On the other hand, the Journal also allows me to subscribe to alert messages the bot will send immediately when there really is important breaking news. Those are great — if the WSJ says I need to know something right now, I probably need to know it right now. I may not need to act, but I’ll understand the impending flood of vague social media posts and offhand comments in Slack about whatever news is going around.

Chatbots are Interactive

This one is last, because while it’s a major difference, it’s one that marketers are still feeling out to see what works. Email, by definition in globally-agreed standards, has to be viewable and reliable everywhere, even on old or unpopular devices. Messenger, WeChat, Kik and Snapchat only need to work for users who’ve installed the app on the phones and computers where it is supported. That has enabled them to add features much more quickly, and without international consensus across corporate and academic giants.

A chatbot can display a photo carousel, a live video, or a 5-question quiz (keep it short) that displays a score at the end. It can initiate a realtime back-and-forth conversation that in email might linger untouched for days, or forever. The possibilities are both exciting and a bit scary.

How to Treat Your Chatbot Differently from Your Mailing List

Read How to Write for a Bot to get started.

Speak to the recipient as a friend

Mailing lists have standardized on newsletter and digital-magazine voices. You wouldn’t email your list, “I would LOVE to hear ur feedback. ❤️ Down?” But that’s exactly what Christina Milian’s chatbot just wrote me. Cool.

Wait until you’ve got a message worth interrupting them

The majority of your recipients will at least glance at your message immediately, because it will pop up on their screens over whatever they’re doing, or in the middle of whatever they’re doing. Is it a must-read-now message? If not, consider waiting.

On the other hand, if you really do need a response or action right now, don’t wait. “We’re going on Conan in ten minutes!” is way better than getting a message about it the next morning. The Dunkin’ Donuts app that sends a coupon for your local store’s daily special right before you leave the house, when you’re thinking about breakfast? That would be awesome in a chatbot.

Don’t broadcast — Interact!

Again, think of that friend who sends you messages but doesn’t seem to be trying to have a conversation with you. Ugh. Email marketers for several of the fourteen websites I had to sign up with for my current job send me email every single day trying to deepen my bond with their products. But they’re one-way conversations: Tips and tricks, features I should try, incentives to share the site with others. Were these sent through Messenger, mingled with my realtime conversations with real people, I’d unfollow them and hope I still got paid.

A chatbot is not a newsletter. A chatbot is not Twitter. A chatbot is not a captive audience by a long shot. You, the serious marketer — I assume, since you’re still reading this — have a chance to connect with an interested, engaged audience a fraction of the size of your mailing list, but potentially far more responsive than the customers who downloaded your app or Liked your Facebook page. You can get a surprising amount of attention from them. But you’ll have to earn it.

Call to Action: Click the 💚 below to tell Medium to recommend this story to other mailing list marketers. 😊

Chatbots vs. Mailing Lists was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Chatbots as campaign vehicles? Rick Boyce was trying to do that in 1996. Now that their time has come, our team of advertising professionals explains how to use chatbots to get your message to the people, and how to get those people to discover your chatbot in the first place.

The Advertising Potential of Chatbots

“With bots we can envision an entirely new high-water mark for brand engagement.”

Bots: A Real Conversation With Customers!

“Fifteen to 60 percent of people who receive a message from a Messenger chatbot with a link to an external URL click the link.”

What if Your Mailing List Had a 60% Click-Through Rate?

How to Use Chatbots for Advertising

“You’ll need to update your 100-slide brand deck and switch the title from ‘digital-first’ strategy to ‘mobile-first’ strategy. And you’ll need to do it fast.”

Conversational Marketing: A New Paradigm for Brands

“A bot that builds a long-term relationship with a customer over time will deliver a much higher return on investment.”

A Must-Read Guide for U.S. Chatbot Marketers

“A bot can create a personal connection, or a Pepsi-level backlash.”

Advertising in Chatbots: How to Do It Right — And Wrong

“We deliberately pushed voters buttons.”

Advertising With Bots: What You Can Learn From Your BFF Trump

How to Advertise Your Chatbot

“To start, you should understand how campaigns on Facebook and Instagratm are structured.”

How To Create The Perfect Facebook Ad Campaign For Your Chatbot

“Even for experienced Facebook advertisers, setting up a Messenger Destination Ad requires a little explanation.”

An Illustrated Guide to Facebook Messenger Destination Ads

Click the 💚 below to tell Medium to recommend this list to others looking to learn about advertising and bots.

8 Must-Read Articles on Bots in Advertising was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Chatbots have an alluring quality: They’re easy to set up. No months of software development, no days of debugging, no hours of trying to figure out how a giant company just changed its API without telling you. But that makes it tempting to rush a bot right out without much, if any, planning.

Management consultant Sriram Sundar published a guide for established organizations that want to roll out a chatbot as a new channel of communication with customers. Much of what he says applies to anyone, though, and Chatbots Magazine has some snappy writeups to help you implement it.

Step 1: Plan To Make the Bot a Full-Time Project

The basic mistake most organizations make, Sriram says, is they assign bot-building to people who already have a full plate of work and responsibilities. Brand consultant Schlomo Rabinowitz has seen companies, as he puts it, delegate crafting trust to an intern. That’s setting up both the intern and the bot to fail.

Before you start — or if you already did and are disappointed with results so far — don’t let the five-minutes ease of launching a chatbot trick you into thinking that planning what the bot will do, how it will do it, and how you will develop and refine it doesn’t require as much serious focus as designing a website or an app, or raising the click-through rate on a mailing list. (Your bot’s click-through rates will be stellar anyway, but you won’t realize how many more customers you’re not engaging.)

Step 2: Identify Your Talent

Handing the bot project to the same folks who did the app isn’t a dumb idea, but there may be a smarter solution. As a startup employee, I ask new coworkers hired for a specific role, “So what else are you good at?”

Is there a screenwriter in the house? A Millennial junior associate who can actually work Snapchat? An engineer for whom flowcharts are second nature? What about that guy who goes home to spend all night promoting his Americana-meets-White-Stripes rock band online? These people may have skills, experience, and a hard-to-define savvy that even a paid consultant might not.

You may hate this, but you’ll have to decide that either the bot is important enough to pull them off their regular roles for a bit, or that you’ll settle for a spare-time project designed by someone who knows how to get Twitter clicks, but doesn’t actually use chatbots. At the very least, your Twitter whiz should be scheduled to put some serious time into figuring out how the very same customer who follows you on Twitter expects a different experience on a chatbot.

Step 3: Think Like a Screenwriter

How to Write For a Bot is a must-read before you design your bot’s conversations with customers. It’s more like writing for a TV series than a mailing list. Here’s a PowerPoint-sized synopsis:

1. Create a character driven identity, even if the character is your brand on a chat channel. Whats its motivation? Its unwritten rules of behavior? Its readily recognizable personal traits? An inconsistent chatbot will either look like a government form, or an amateur mess. Poncho the weather bot has multiple screenwriters with comedy experience, who’ve created a weather correspondent with an engaging sense of humor.
2. Speak in your brand’s recognized voice. Do you have catchphrases? Well-known ad slogans or characters? Lyrics from your hit songs? Christina Milian’s bot takes a second to look like it’s typing before its response appears, because that’s what chatting with Christina herself is. Don’t talk like your media kit in a chat session. Every time I check out a favorite rock band’s bot that tells me to “Follow us for exclusive content,” I remember why everyone hates the music industry.
3. Always speak in a one-to-one manner. A chatbot session is not Twitter. It is not a mailing list. It is not a Facebook page, even if it’s on Messenger. Broadcast-type messages meant for a crowd read like junk mail in a chat session.

Hard Planning Makes It Look Easy

If it seems too easy for me to tell you to stop, plan, and allocate resources for your bot, it’s not my armchair opinion. I do it myself for these posts. I’ve been writing professionally for twenty years. Readers often presume that by now I just type up posts as glibly as America’s president posts tweets. But even after thousands of articles, I stop before I start typing into Medium, and make an outline of what exactly I want to tell you, and what wording or images I want to use. I block out how much space I want to devote to each section or idea. I consider how it fits into the site’s other content, and where it needs to be careful not to read as if it were written by Octane AI’s marketing department, since it’s no secret Chatbots Magazine is published by Octane AI’s marketing department.

I spent four months — longer than I’d planned — coming up to speed on chatbots, NLP, AI, and chatbot applications before I began glibly handing out advice. I built some bots that other people actually use. And yes, I’ve created a character for Chatbots Magazine, a subset of myself that sticks to the topics readers are here for, and has specific levels of reading level and technical talk. Paul the CBM writer has more you-can-do-it cheer and less of the deadpan sarcasm I’m known for elsewhere. And after writing a draft, I review it to ensure you don’t need to get my jokes, or even get that the jokes are jokes, to leave with the right takeaway. My manager metaphorically summed it up when I started: “Take some time to train your bot.”

Why I’m telling you this: Typing into Medium takes the least time of my post-writing process. By contrast, I’ve watched enthusiastic bot makers leap right into the configuration screen and start serving a beta bot to customers minutes after they’ve created the necessary accounts. That’s great. That’s good Internet thinking. Ready, fire, aim! Move fast and break things! Something about disruption! You fearlessly launched your first bot, good for you.

But now, sit down and, as our contributing management consultant advises, plan to have a plan.

Click the 💚 below to recommend this story to other Medium readers eager to launch a bot because it’s super-easy, right? 🙂

3 Stop-and-Plan Steps Before You Launch a Bot was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.