Get ready for the bot revolution

Will bots liberate IT from mundane tasks, or will their development and upkeep only add to IT’s workload?

Get Ready for the Bot Revolution - illustration by Richard_Borge [SINGLE USE/Computerworld]
Richard_Borge

For today's overworked, time-strapped IT employees, bots are more than simply apps that perform automated tasks, like delivering weather reports or taking pizza orders. Rather, they're a respite from endless help-desk calls, constant software updates and tedious server maintenance jobs.

"Eighty percent of IT's effort is focused on mundane, grunt work — ditch-digging to keep the lights on with barely 20% spent on innovation," says Frank Casale, founder of the Institute for Robotic Process Automation (IRPA). But bots promise to change all that, he says, by "taking on the bulk of the routine, dismal work that makes IT workers feel like human robots."

Computerworld illustration by Richard Borge - Get Ready for the Bot Revolution [2016/single use] Richard Borge

For example, AT&T is using bots to automate humdrum data-entry activities. And 1-800 Flowers has rolled out bots to help customers place online orders, while TV network CNN uses bots to deliver breaking news and personalized stories. By handling tasks that are either directly overseen by IT, or supported by IT resources, bots are fast becoming "magical for most IT departments," says Casale.

Bots are hot

Bot developers who need help pulling the proverbial rabbit out of the hat can turn to new technologies from Facebook, Microsoft and an emerging crop of bot-centric businesses. In April, Facebook announced the launch of Bot Framework, a platform that allows developers to build chatbots for use on a variety of messaging platforms, including Facebook Messenger, Slack, Skype and WeChat. Since then, more than 11,000 bots have been created.

In March, Microsoft unveiled Microsoft Bot Framework — a set of tools for creating chatbots on multiple platforms, including Skype, Slack and Telegram. And an increasing number of startups, such as Pandorabots, Rebot.me, Imperson and Reply.ai, are releasing third-party platforms for people who want to build enterprise bots with a distinctly human touch.

In fact, according to a recent report published by Transparency Market Research, the global IT robotic automation market is expected to grow to $4.98 billion by 2020 — a 60.5% leap from 2014.

It's easy to understand why: Thanks to powerful platforms, you can develop a bot in about one-fourth the time it takes to build a standard mobile app. And because bots don't rely on costly servers, they're approximately 50% cheaper to build and maintain, according to Casale.

But for all the optimism about bots liberating IT from mundane tasks, there's plenty of concern regarding the impact of bots on overall IT workloads.

To be sure, bots promise to free IT workers for more important tasks by automating activities such as ticket management, server load balancing and customer service. However, as an increasing number of companies embrace bots as an easier and cheaper alternative to web apps, many IT professionals are questioning whether bots will create extra work for already beleaguered IT teams.

Get Ready for the Bot Revolution - illustration by Richard_Borge [SINGLE USE/Computerworld] Richard Borge

There are plenty of labor-intensive byproducts of the bot revolution. The list includes honing bot development skills, identifying new security vulnerabilities and addressing bot design flaws.

While there's no definitive solution for challenges such as those, savvy IT leaders are determining how best to embrace bots and tackling the challenges that come with these much-hyped app alternatives.

A prescription for success

At HealthTap, the decision to build a bot was a no-brainer because the healthcare company has an ace internal IT team and a robust operating system. HealthTap offers a mobile app that lets consumers access a network of more than 100,000 doctors anytime, anywhere via secure video or text chat. Visitors can ask questions about everything from pregnancy to palpitations and receive personalized answers.

So when Facebook announced its bot platform for Messenger, HealthTap was eager to take advantage of the social network's vast audience reach. The result is a HealthTap chatbot that allows users to type a question into Messenger to receive free responses from doctors and view similar queries. HealthTap built the bot in "a few short weeks," says Sean Mehra, the company's head of product. But the quick turnaround belies the truth behind bot development, he adds, noting that building the bot that fast was a phenomenal feat.

"If you want to launch a smart bot that handles a complex use case, that's not a trivial exercise," warns Mehra. "The ability for us to take any human language, parse it, understand it, figure out what health concepts pertain to it, map it to a library of answers and send it to the appropriate specialist has nothing to do with a lack of complexity."

Instead, Mehra says HealthTap's bot success is "really a testament to the fact that we have built a platform that's very developer-friendly."

That platform is Health Operating System (Hopes), a proprietary tool that serves as the operating system for HealthTap's mobile app. By "reusing a lot of the infrastructure created when building the operating system," Mehra says, HealthTap was able to design and deploy a full-fledged bot quickly while avoiding any "heavy lifting."

Another key competitive advantage for HealthTap is a workforce that would fit right in at a Silicon Valley startup. The collective expertise of the company's 100 employees encompasses product design, scalable back-end systems, linguistics, natural-language processing, machine learning and big data. Mehra says this combination of "both core technology expertise on the engineering side and a really strong understanding of user experience with a product and design focus" helped HealthTap get its bot off the ground.

Tapping outside IT talent

But not all companies have the same IT resources as HealthTap. When it came time for Transcosmos America to design and develop a bot to provide customers with fast answers to their most pressing questions, the master U.S. distributor for PC manufacturer VAIO turned to Reply.ai.

Reply.ai helps businesses build and manage chat bots over an array of messaging platforms. Tom Coshow, managing director for Transcosmos America, says that while there are a growing number of vendors offering bot development services, Reply.ai's "agent takeover" capability set it apart. This feature allows a human to take control of a conversation when a bot can no longer sufficiently address a customer's needs.

Get Ready for the Bot Revolution - illustration by Richard_Borge [SINGLE USE/Computerworld] Richard Borge

"For a high-end brand like VAIO, we never wanted to put our consumers in a situation where they would be trapped inside a bot that couldn't answer their question," says Coshow. "So agent takeover was super important to us. The idea that you could type 'agent' at any time and get to a live person was a big deal."

To make that possible, Reply.ai needed to integrate its agent takeover feature into the contact system that runs throughout the VAIO call center. After devising a number of creative workarounds, Coshow says, Reply.ai eventually discovered a way to "drop messages" into the contact system so that bot conversations are automatically "routed to an agent inside the normal workflow of a call center."

That not only creates an uninterrupted flow of communication between company and customer, but also enables continual improvement of the VAIO bot's functionality without requiring IT teams to monitor performance and flag glitches.

"The beauty of what we built with Reply.ai is that when a customer comes out of the bot, and drops down to the agent, the whole bot customer transcript comes with them," says Coshow. "So not only does the agent know what the consumer was experiencing in the bot, but he sees exactly where the bot failed."

Ease of use is another feature that drew Transcosmos to the Reply.ai platform. "Any power user can build a bot on Reply.ai," says Coshow, who worked as a programmer earlier in his career.

Price also played a key part in the vendor selection process: Coshow estimates that it cost $25,000 to design and develop the bot — a fraction of the cost of working with high-priced tech consultants.

"I've heard of agencies charging $150,000 to build a bot," he says. "To me, the way to build a bot is to build it with help from your customer service team. I don't want to build bots with two programmers locked in an office. You want to build a bot in an organic way with your customers."

Virtual marketing director

Some businesses are even discovering bots that replace certain IT and marketing roles. Take Noli Yoga, for example. The yoga apparel and active wear startup needed an easy and cost-effective way to run targeted ads on Facebook.

As a newly minted entrepreneur, Noli Yoga founder Slava Furman was already handling everything from design to fulfillment. And being an attorney by trade, he didn't know a thing about building a mobile presence or crunching data to figure out target audiences. That is, until he discovered a marketing bot called Kit in the Shopify app store. For $50 per month, Kit sends Furman daily text messages that automatically notify him of profits, suggest Facebook ad placements, encourage budget increases for previously successful campaigns and flag slow-moving products.

"It's an amazing application for someone who doesn't have the experience or time [to manage an online store]," says Furman. "It definitely saved me a ton of time and a ton of money, especially in the beginning." In fact, since launching the site in May 2015, Furman has tried out two different agencies to help manage the fledgling company's Facebook ads at a cost of nearly $4,000 per month for less-than-stellar sales results.

Conversely, Furman says, Kit was responsible for 80% to 90% of the $1.2 million in sales Noli Yoga garnered in its first year of operation.

The imitation game

Whether your bot strategy involves pinning all your hopes on internal IT, tapping the IT expertise of a third-party provider or creating a virtual IT manager, obstacles still abound for most businesses.

"When it comes to bots, what's most challenging for people is having a bot pass the Turing test," says HealthTap's Mehra, referring to a test that measures a machine's ability to imitate human behavior and intelligence. "How do you mimic a human conversation in a way that doesn't feel mechanical and robotic?" Mehra asks. "That's an area that the entire industry will focus on in the coming years. And it's not a simple problem."

For there to be any improvement, Mehra says we'll need to see significant advances in our understanding and application of linguistics, natural-language processing and artificial intelligence.

In the meantime, Coshow of Transcosmos says the right content and tone of a text message can help compensate for a conversation's lack of human-like subtleties.

"You need to select verbiage that matches your audience," says Coshow. "If I'm trying to provide customer support to some executive in an airport lounge, and my bot is being flippant, that's a disastrous customer service experience," he says. "When you're representing your brand and providing customer support, it's important to err on the conservative side."

A saving grace is the presence of a living, breathing human being. Futuristic talk about machines replacing people notwithstanding, humans are still a critical part of the bot equation. "A big deal about bots is going to be whether or not you have access to [a human] agent," says Coshow. "Bots are going to fail, and when they fail, you need to be able to provide a live person. Otherwise, it'll be a terrible consumer experience."

But are they safe?

One of the ways bots can fail is by opening up security holes.

A poorly designed bot can expose a great deal of information about a customer or a software program, says Yegor Buga­yenko, a software architect who helped create Rultor, a chatbot that helps programmers build and compile their projects on Github and now boasts more than 400 users.

Before releasing a bot for public use, Bugayenko says, IT teams need to carefully review code and ensure that sensitive information is protected. Data encryption and stringent access control policies are crucial, he says.

IT professionals who take the time to sharpen their bot design and development skills can improve their chances of avoiding security snafus. After all, says Bugayenko, building a bot "is not a typical skill set. It's not so easy for a regular programmer to design a chatbot."

Setting standards and boosting bot smarts

IT maintenance is another issue that bot developers need to consider. There are currently no industry standards governing bot design and development — a predicament that might make it challenging to integrate bots as an increasing number of them are deployed across disparate messaging platforms. Nevertheless, Casale predicts that, "over time, we will see standards in place as it behooves the marketplace."

Others are banking on more innovative technologies, such as machine learning, to reduce IT maintenance tasks. That's HealthTap's approach. Mehra says that the company's bot becomes increasingly smarter with each and every question asked and answered.

"We have an army of doctors who are using the system and essentially training that machine," says Mehra. "So when doctors review and answer questions, they can see if a response was improperly tagged or rerouted, and they can then fix it." Ultimately, he says, this combination of "crowdsourcing and an algorithmic technology is much more skillful" than requiring IT staffers to constantly tweak and update a bot.

Looking ahead

User experience, integration, security, maintenance — bot-related activities like those are all likely to put more work, not less, on the plates of IT teams. Development of bot-related skills, features such as agent takeover and the emergence of industry standards could ease the burden. But one thing is certain: There's no way to automate bot success.

IT leaders must play an integral role in the process of properly designing, developing, deploying, testing and improving today's bots. Even Furman, who has had so much success with the Kit bot at Noli Yoga, admits that "once you get to a certain point [with a bot], it kind of tails off. There is a little bit of a limit or ceiling there."

While breaking that ceiling might create more work for IT in the short run, the hope is that turning mundane tasks over to bots will ultimately give time-strapped techies the freedom to pursue more pressing projects.

Copyright © 2016 IDG Communications, Inc.

Bing’s AI chatbot came to work for me. I had to fire it.
Shop Tech Products at Amazon