Latest news

Chatbots for Business: Needs, Misconceptions, and Marketing Hype

No comments

There’s a lot of confusion surrounding chatbots, machine learning, and artificial intelligence. Inconsistent terminology, unrealistic expectations from science fiction, and marketing hype from vendors all share part of the blame. Misunderstandings, misdirections, and miscommunication are preventing widespread adoption of these content delivery channels. I’d like to help us all get on the same page—or at least—reading from the same hymnal.

Since 2001, myself and other technical communication professionals and content strategists, have been evangelizing the need for organizations to adopt advanced information management practices and tools. We work to convince organizations to move from the creation of hand-crafted deliverables like documents and web pages and toward the creation of intelligent content—modular, semantically-rich, format-free, structured content designed to be managed, translated, assembled, and delivered (automatically) by machines.

We developed this mindset—and spawned the creation of best practices, content standards, and an entire software industry sector—out of necessity.

In the early years of this century, as we struggled to produce software documentation for use by customers, we recognized the need for a more scalable approach. In order to deliver role- and task-based content across multiple operating systems, in multiple delivery formats, targeted to multiple audience segments, translated into multiple languages, and delivered to multiple device types, we realized we could no longer rely on traditional content creation and management approaches. That recognition led us to adopt a new approach (single source publishing) that leveraged lean manufacturing processes relied on a unified content strategy designed to help us efficiently create content that is both human-readable and machine-processable.

See: Unified Content Strategy: Fact or Fiction?

By creating intelligent content, we were able to future-proof our production methods; affording us a readiness to more easily support new channels of content delivery as they come online. Organizations that have adopted these advanced approaches can quickly adapt their content framework to deliver content at scale to emerging delivery channels like augmented reality, chatbots and conversational interfaces like Amazon Alexa, Google Home, Microsoft Cortana, and Apple’s Siri.

But, despite building a future-proof infrastructure designed to support emerging content delivery channels, there are still a few big challenges to overcome.

Success is not guaranteed; Creating intelligent conversational content is a big part of getting it right

Chatbots—and their voice-enabled cousins—are conversational interfaces. They require conversational content—content designed to be served up in discreet conversational units, sometimes called question/answer pairs or thread-title/reply. When thoughtfully implemented, conversational interfaces can answer questions—and respond to commands—from consumers.

Try this: Ask Amazon Alexa, “Who is Scott Abel?”

Don’t have an Amazon Echo device. No problem. Fire up the Amazon app on your smartphone—or head to the Alexa skill testing tool from your web browser. Press the microphone icon and ask the same question.

In most cases, chatbots interact in a conversational manner with consumers, often playing the role of a digital content tour guide, helping to direct us toward the content we need. Poorly implemented, they frustrate, confuse, and bore consumers.

Done well, conversational content experiences can drive engagement and satisfy prospective (and existing) customers by making it easy for them to discover useful, relevant content using natural language queries, instead of unnatural commands conjured up by software developers.

Done exceedingly well, chatbots can be useful in not only guiding us toward answers we seek, they also can be instructed to do work on our behalf, like the chatbot that can setup a website for you (like this one; see below) in less than 2 minutes. That’s useful.

The framework of was created by a chatbot.—a chatbot that sets up a website landing page by asking you questions—helped me set this landing page up in less than 2 minutes. Within an hour, I had the site up, connected to PayPal and started earning revenue selling tickets.

There are caveats—and dangers—to moving to conversational content, but the benefits are likely to outweigh the drawbacks in the long run.

Unfortunately, the vast majority of our content is not ready for conversational prime time. In short, our written words were not designed to be delivered conversationally. We never envisioned that the information we craft would be delivered in chunks by an automated system or that the content we create to serve existing customers would end up being read aloud by a voice-enabled content delivery device.

Try this: Command Google Home, “Books by Scott Abel”

Then, try that same command, but insert a different author’s name. Try it several times with different names. Some books are discoverable by Google Home, others are not. Why?

As such, our content is ill-prepared for the world of delivery options consumers have available to them today. Our content was written with consumption by human consumers in mind; but, not computers. That approach is no longer sufficient.

In a world where computers are used to seek out, process, and serve-up answers to questions on demand 24/7, our content needs to be intelligent. It needs to be prepared so that it is both machine-processable and human-consumable. To provide maximum value, much of our content needs to be constructed in ways that provide computer systemsmwith the semantic cues they need to deliver the right piece of content to the right person, at the right time, in the right language and format, on the device(s) of that persons’ choosing.

What is intelligent content?

Intelligent content is content which is not limited to one purpose, technology or output.

It’s content that is structurally rich and semantically aware, and is therefore discoverable, reusable, reconfigurable and adaptable. It’s content that helps you and your customers get the job done, often automatically.

See: What is intelligent content?
See: How to optimize for voice search 
Watch: How Structured Content Makes Chatbtos Helpful (recorded webinar)
Watch: Building Chatbots with Intelligent Content (recorded webinar)
See also:
Personal Assistants and Voice Search Optimization

Of course, some of our most important content will need to be rewritten in a conversational tone, so it can be delivered to humans by talking machines in a way that will seem more human, more natural. Our goal should be to provide content that makes sense to consumers whenever and wherever they need it. When they ask a question, we should be able to provide answers that make sense for the content delivery channel they choose to use.

See: Tips for writing in a conversational tone

Chatbots and conversational interfaces are not artificially intelligent—not even close

There’s a lot of confusion about the role of artificial intelligence in chatbots and voice interfaces. Many people incorrectly assume that chatbots and voice interfaces leverage artificial intelligence (AI). More often than not, that’s not the case. If you hear a sales pitch from a chatbot-maker that sounds like your foray into conversational content will be super easy and amazingly magical—run. Run away from that person and never look back. Unsubscribe from their newsfeed. Block their telephone number on your phone. And create an automatic rule to route their emails to the SPAM folder, pronto.

As is predictable with any new—and promise-filled—technological advancement, there is a lot of exuberant exaggeration (and sometimes, pants-on-fire lies). It’s not surprising, then, that unrealistic expectations set up many chatbot purchasers for disappointment. But, vendors and consultants are responsible for some of the confusion and much of the pants-on-fire-type lies.

For example, chatbot maker Inbenta, provides a “guide to using artificial intelligence powered chatbots as a way of reducing costs and gaining professional advantage.” In chapter 1 of their Chatbots for Business Guide, the very first words are a great example of the type of content challenge consumers face.

“A chatbot is a computer program that mimics conversation with people using artificial intelligence.”

Of course, that’s a false statement. It’s a flimsy-truth, not quite a lie, but definitely not a statement of fact.

Later in the same marketing piece, Inbenta goes on to explain that in order to get the benefits they describe—a reduction in support tickets, automated delivery of answers to commonly asked questions—chatbots can be integrated with knowledgebases and machine learning platforms. “They can also be programmed to triage your helpdesk and contact center tickets…”

The operative word here is programmed, not artificial intelligence.

See: There is No AI without IA

Chatbots should augment our work and allow us to focus on creating agentive content solutions

Chatbots and voice interfaces are not magical technological wonders capable of mimicking cognitive functions such as thinking, intuiting, and problem-solving—at least not yet (and not without significant time, expense, and effort).

Instead, they are almost always used to encourage menu-driven conversations—conversations governed by good old-fashioned rules. If this, then that logic that relies on conditions and triggers powers their scripted behavior. And, it is a script. A guided, programmed, scripted conversation. Biomimicry in action; not artificial intelligence.

Much of the time, what passes for artificial intelligence is nothing more than automation with a twist. In the rare cases where chatbots are capable of doing more than just following the rules, it’s more likely that they borrow extra capabilities from machine learning applications like natural language processing or pattern recognition. There’s nothing wrong with this approach, but there’s also nothing remotely close to artificial intelligence or cognitive computing going on here.

See: MilaBot: A Deep Reinforcement Learning Chatbot

As Vaisagh Viswanathan writes in his Chatbots Magazine column, How To Make A Chatbot Intelligent:

“When information is stored in the correct way with the right rules and data structures, it can be immensely powerful, even multiplying the effectiveness of any learning that is done.”

Our role is to create our content in way that makes it both consumable by humans and understandable—and actionable—by machines.

Recorded webinar: Building Chatbots with Intelligent Content

AI-washing: Don’t believe the marketing mega-hype

Some researchers say artificial intelligence is achieved when a machine mimics cognitive functions such as learning and problem solving. This human-like behavior can be measured by testing, oftentimes by using the Turing Test. The test is used to determine “ a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human.”

Stanford University computer science researcher and Professor Emeritus, John McCarthy extends the definition of artificial intelligence by pointing out that it is also “the science and engineering of making intelligent machines, especially intelligent computer programs.” He says that “a machine that passes the (Turing) test should certainly be considered intelligent, but a machine could still be considered intelligent without knowing enough about humans to imitate a human.”

“It turns out that some people are easily led into believing that a rather dumb program is intelligent,” McCarthy says.

Unfortunately, he’s right. People are often easily bedazzled by hype. The allure of artificial intelligence is strong. To safely move toward more advanced content solutions we have to be vigilant — and be on the lookout for unscrupulous vendors who seek to attract us with AI-washing, what Gartner refers to as a “confusing” overhype; an effort at rebranding software products as “AI-powered” or “AI-enabled” in order to capitalize (make sales) on the excitement surrounding artificial intelligence.

In a recent report entitled, “How Enterprise Software Providers Should (and Should Not) Exploit the AI Disruption,” Jim Hare, Vice President of Research at Gartner said that such efforts by software makers “are not helpful, because AI-washing often contains nothing more than empty promises.”

“Many technology vendors,” Hare says, “are now ‘AI-washing’ by applying the AI label too indiscriminately.”

Terminology, marketing mumbo-jumbo—and hyperbole-filled use cases—aside, it’s important that we focus on the problems we’re trying to solve, not on trying to make solutions fit into the tools that vendors bring to market. The best tool is the right one for the job.

When attempting to overcome a challenge, it’s a well-established best practice to declare your goal, identify the problems, and then work through the process of deciding the best way to solve them. Oftentimes, the improvements we’d like to make to content-driven customer experiences can be accomplished using existing techniques and technologoies. Chatbots, voice interfaces, and machine learning may—or may not—be needed.

Regardless of what vendors might have you believe, artificial intelligence is not likely the solution to every content experience problem.

See: Your bot will not pass these four simple tests, so why bother with artificial intelligence
Read: 4 Uses for Chatbots in the Enterprise

100,000 Facebook chatbots: The problem with big numbers and success

Michael Lam, Founder of Custom Bot Design, says Facebook (and the technology press) are partially to blame for hype.

“Maybe you’ve heard this statistic. ‘There are over 100K chatbots on Facebook Messenger alone,’ Journalists and bloggers write about this number — and presenters include it in their conference presentations—as though it’s some type of validation,” Lam says.

“Unfortunately, chatbots success is NOT based on the quantity of chatbots created, but rather, the quality and utility of Facebook chatbots, which are today—at best—poor.

Facebook is to blame and they know it. They oversold the promise of chatbots and artificial intelligence, and now they have to backtrack their way out of that mistake.”

After being criticized by the technology media for their ambitious, yet error-filled launch of chatbots (see: frustrating and useless and 70% failure rate) Facebook seemed to place most of the blame for their less-than-stellar chatbot launch on their users.

“We never called them chatbots,” Messenger vice president, David Marcus told a group of reporters in April 2017 at the F8 Conference.

“We called them bots. People took it too literally in the first three months that the future is going to be conversational.”

Getting started: Preparing content for chatbots and voice interfaces

Where can you learn more about the issues impacting the way we create, manage, translate, and deliver content to chatbots and voice interfaces?

One of the best resources is Information Development World (IDW), a conference I co-produce with the folks at Content Rules, focused on serving the needs of content production teams working for the world’s biggest brands. The event specializes in helping attendees create machine-ready content optimized to be adapted, personalized, and delivered dynamically.

IDW is the conference for technical, marketing, and product information developers — the folks responsible for creating exceptional customer experiences with content. IDW is a twice-yearly gathering of professional product content creators, strategists, managers, translators, engineers, marketers, trainers, editors, and delivery specialists designed to help companies that value content as a business asset produce content that is engineered for both humans and machines.

The conference takes place at the Quadrus Center in Menlo Park, CA, November 28–30, 2017.

The three-day event features presenters from Intuit, Amazon, Facebook, Microsoft, and hands-on mini workshops including:

The conference is different than most in that it’s designed to ensure attendees have a full understanding of the topics that matter. Instead of attendees choosing from sessions in multiple tracks, IDW brings conference-goers together in one big room. Presenters switch. The audience does not. This assures that everyone learns the same things—and that attendees leave with a common understanding: what are chatbots?…how do they work?…why should I care?…what do I do first if I do care?

Check out the roster: Information Development World

Required reading: Christopher Noessel (IBM) provides food-for-thought to anyone aiming to leverage machine learning in practical ways to improve customer experiences. Highly recommended.

The future of chatbots is agentive

“Whether you write a UI or a chatbot, it doesn’t really matter. The user will come to your solution only if it saves time.” — Shival Gupta

Despite the challenges of communicating the truth (and not being sucked in by the hype), chatbots and voice interfaces are the beginning of a new period of customer-focused improvements in the delivery of content experiences. They are already part of our present. They will be part of our future.

For as long as I can recall, we’ve been focused on creating content that helps our customers do work. But, over the next decade, I believe we should focus instead on creating content that does work for our customers. It’s the evolution of content. And, it’s what consumers want.

That’s why Val Swisher and I created Information Development World and decided to focus the next few years of our conference on chatbots, voice interfaces, machine learning, and agentive technology.

Highly Recommended Reading: Designing Agentive Technology: AI That Works for People, by Christopher Noessel, IBM (Rosenfeld Media)

We aim to provide content of value that helps content creators, managers, and other content industry professionals understand how these fast-growing delivery channels can help us achieve our business goals. We’ll be focusing on helping companies that understand the importance of delivering the right content, when, where, and how their prospects and customers desire.

Our goal is to help big brands move away from creating content that helps us do work, toward content that does work for us.

Care to join us?

magnoliaChatbots for Business: Needs, Misconceptions, and Marketing Hype

Related Posts

Leave a Reply