No robots required: AI will eliminate these jobs first
From Alexa to Viv, the world is now full of voice-enabled cloud-connected assistants. But Amelia is more than merely a series of speech-savvy algorithms -- she's Siri with a doctorate in psychology.
Thanks to advances in semantic analysis, Amelia can step you all the way through a sophisticated business process, like purchasing insurance. By analyzing the language you use and your tone of voice, Amelia can also detect when you're unhappy and pass you immediately to a human. Then, as you're unspooling your anger to a long-suffering support agent, Amelia listens and applies what she learns to future interactions.
"Amelia is not just another fun and friendly chatbot," says Ben Case, solutions architect for IPsoft, the “digital labor company” that created her. "Her goal is to be practical and pragmatic; to answer questions, retrieve information, and solve problems using her semantic network and sophisticated sentiment analysis."
Today, Amelia is being trained to help sell insurance, prequalify people for mortgages, and provide customer support. While most of these efforts remain pilot programs at the time of this writing, Amelia is starting to emerge into the business world. In May, Accenture launched a dedicated Amelia practice to help its clients deploy virtual agents across their organizations.
One day, Amelia or one of her artificially intelligent cousins might become your indispensable IT assistant, taking on the work of level-one tech support, troubleshooting problems, or detecting security anomalies. Then again, she might replace you entirely -- or you could end up working for her.
Here, we take a look at AI’s evolving role in the IT workplace, with an eye on how it might impact your career in the years ahead.
Everybody thinks they know what AI is. It's the thing that enables computers to defeat chess champions and win at "Jeopardy." It's what allows Siri and Alexa to understand what we're saying to them and respond in a humanlike manner.
When exposed to the dregs of the Internet, AI can also be turned into a Nazi-loving hate-spouting sex bot. And thanks to Hollywood, we know AI as the thing that will eventually attempt to enslave and/or exterminate humanity.
The classic notion of AI is a machine so smart it fools you into thinking it's a person. But "artificial intelligence" has become a catchall for a jumble of technologies -- such as machine learning, natural language processing, cognitive computing, and robotic process automation -- that automate rote tasks and help people make better decisions.
"AI doesn't have to pretend to be a person to have a huge value to the world," says Scott Crowder, CTO and vice president of technical strategy and transformation at IBM Systems. "It's about providing information and insight to humans, so we can do a better job."
That's one reason why IBM prefers the term "intelligence augmentation" -- IA, not AI -- and defines its "Jeopardy" champion Watson supercomputer as "a cognitive computing technology that extends and amplifies human intelligence, working in partnership with professionals."
AI is already serving on the front lines of service and support via voice-enabled virtual customer agents like Amelia. But because it also excels at analyzing massive amounts of unstructured data, the technology is ideally suited for identifying potential security threats or helping drive business decisions.
That -- theoretically -- can free up IT professionals to spend time on higher-level tasks.
"Fundamentally, AI changes the business and operational dynamics in any industry by enabling machines to find answers and make decisions that humans make," says Tim Tuttle, founder and CEO of Mindmeld, makers of a conversational AI platform. "For example, AI can help field a much larger range of problems and answer those not requiring a person's time, giving IT people more time to focus on the difficult questions machines are not equipped to answer.”
Early forms of AI, like robotic process automation, will have its greatest impact on IT, says Rob Brindley, a director at consulting firm Information Services Group. RPA will be deployed to automate mundane and highly repeatable tasks, such as monitoring systems, distributing software, rerouting workloads, support, and provisioning.
That process is already well underway. For the past eight years, IPsoft has been applying machine learning to problems like remote management of networks and other IT infrastructure. The company has created more than 20,000 "virtual engineers" -- AI-driven processes that can diagnose common problems and apply known solutions -- for clients like Cisco and IBM.
IPsoft's virtual engineers can handle nearly 60 percent of all incidents with no human intervention, says Jonathan Crane, chief commercial officer at IPsoft. When a virtual engineer runs into a problem it can't solve, it escalates to a human specialist who's trained in that particular issue.
Eventually, some low-level IT functions, such as problem management or provisioning of new employees, could be handled by an AI-based front end like Amelia, says Crane.
"You'd tell her you need equipment for three people you just hired, and she could look up their information in the personnel database, figure out which equipment and security codes they need, and configure them appropriately," he says. "This is where you really start to save money and speed up the management of your IT environment."
While automation like this will likely replace employees on the lower end of the IT spectrum, these are jobs that have largely been outsourced or offshored already, Crane adds.
"Front-line support people typically escalate 80 to 90 percent of the problems they encounter," he says. "Why not use computer diagnostics for that process"
Another area where AI will change the lives of IT pros is its ability to sift through petabytes of data, identifying patterns and looking for anomalies.
"Some IT departments have to examine billions of actions a day from log files or streaming events, such as network security on connections," says Nik Rouda, senior analyst at Enterprise Strategy Group. "Humans can look at perhaps a few hundred a day. AI will be able to learn what’s suspicious, flag it, and even respond."
That makes it a prime candidate for identifying zero-day vulnerabilities and defending against botnet attacks, says Carl Herberger, vice president of security solutions at Radware, a cloud-based security service.
“Most of the major security threats -- like application DDoS, brute force, and SQL injection -- are executed at least in part through botnets designed to select actions based upon the anticipated responses from a human defender," says Herberger. "As people have become increasingly predictable in detection and mitigation, the bad guys are designing tools to adjust to our defenses faster than we can detect their changes. AI and automation will be critical in fighting back.”
But with the number and seriousness of attacks growing daily -- and an estimated 209,000 openings for cybersecurity professionals yet unfilled -- AI would serve to supplement security officers, not replace them, says Brad Lovering, co-founder and chief engineering officer at IT security firm SignalSense.
"I don’t think anyone working in cybersecurity right now is worried about losing their job to AI, because expert human insight is still so highly valued and the field is so drastically understaffed," says Lovering. "But I certainly see a future where AI can sense, signal, and squash these issues with little or no human intervention."
Likewise, gathering and analyzing massive amounts of data from disparate sources is a prime candidate for applying intelligent algorithms. That's why AI will play a big role in assuring compliance for companies in highly regulated industries, like health care or finance.
"Armies of people are necessary today to handle the increased regulatory pressures the financial industry is facing," says Dan Adamson, CEO of OutsideIQ, which develops AI solutions for risk assessment. "Those processes can be done in a more auditable and consistent manner through AI."
For example, under Dodd-Frank, financial services firms may be asked by regulators to "play back" a trade made months or years in the past. Traditionally, compliance officers would have to manually sift through communications data from a wide range of sources -- email archives, phone records, chat logs, document management systems, and social media posts, as well as various trading systems -- to re-create the transaction.
"Simply trying to normalize all that data and place it on a timeline is extraordinarily difficult and time consuming," says Harald Collet, global head of Bloomberg Vault, an information management and analytics service that uses machine learning. "It's a scenario where it might have taken two months to respond to a regulator, and now the regulators are saying you have to respond within 72 hours."
Collet says more than 1,000 financial firms have deployed Bloomberg Vault so that they can monitor and archive trading activities and respond to regulators quickly. But the larger goal is to use machine learning for predictive analytics, so Vault users can identify potential issues before they become regulatory headaches.
"Firms want to detect patterns across large data sets and flag irregular trading patterns," he says. "For example, they want to be able to see chats where a trader has given a price quote to the customer, then immediately compare that quote to the actual price in the market at the time. The ability to see the connections between price points, chat, and trading patterns is what our clients are looking for."
Over the long term, systems like Bloomberg Vault can help optimize business processes and identify new opportunities, says Collet. As AI drives more decisions, the role of technologists in the organization will also change, notes Steven Hillion, chief product officer at Alpine Data, an advanced analytics platform for enterprises.
"IT's role is moving from one of data governance and security to one where they think about how data is delivered, how data provides insights into business operations, and how to operationalize those analytics into strategies that will positively impact the company," Hillion says.
It's been true since the industrial revolution: As technology replaces low-level workers, new jobs are created to service and improve that technology. AI is no different.
"We live in a world where machines are going to continue to squeeze out humans," says Anant Jhingran, vice president and CTO of Apigee, a maker of intelligent API platforms. "That's definitely going to happen."
But AI is also likely to create new jobs. IPsoft's Crane says organizations may end up hiring digital compliance officers who can make sure systems like Amelia are making the right decisions, especially when it comes to regulatory issues. Organizations may also need a digital marketing officer who can come up with new ways to respond when Amelia detects that a customer is delighted and primed to spend money, adds Crane.
AI will put a lot more pressure on IT infrastructure, especially in the areas of data storage, memory, and compute power, says IBM's Crowder. But it'll also require existing tech pros to master new skills, such as the ability to integrate intelligent voice-enabled user interfaces into existing applications.
"There will be a huge demand for specialists who can 'teach' bots by designing for conversational UX," says Ilya Gelfenbeyn, CEO of Api.ai, a platform providing tools for building speech and text interfaces. "These bots will replace a large number of GUIs and act as interfaces for devices without screens, especially for the Internet of things."
Ultimately, the secret to survival in this brave new world is bringing AI to everything you create, says Jhingran.
"The more intelligence you build into your code, the more secure your job will be," he says. "If you bring intelligence into all code you write, you won't just be secure in your job, you'll be a superhero."