Where Is AI Headed in the Enterprise?
Updated · Nov 08, 2016
Talk about artificial intelligence, and most people think about machines that can reason. In a business context that would mean clever AIs that could use their powerful intellects to become formidable business leaders that could outwit mere humans.
But since the late 1980s, those working in the field of artificial intelligence have shifted away from reasoning because it turned out to be far harder than people had anticipated. IBM’s Watson notwithstanding, the focus has been more on machine learning and pattern recognition — activities which, like reasoning, also take place in the human brain.
AI and Pattern Recognition
Pattern recognition is what enables people — and artificial intelligences — to understand that a photo and a different line drawing may both be representations of, say, a cat.
Huge advances in software-based pattern recognition have taken place over the last 20 years, allowing computers to recognize patterns in data, images, words and phrases. “Machines are good at sorting things into buckets using machine learning but planning, formulating hypotheses and proofs? That has proved much harder,” explained Nova Spivak, a technology futurist and serial entrepreneur. “Reasoning is different. Very little progress has been made on that front.”
One possible explanation for the advance in pattern recognition capabilities is that the processing power available to carry it out has become much cheaper. Low cost graphics processing units (GPUs) that can do many processing tasks simultaneously are ideal for pattern recognition work. Between 2008 and 2016, there was a million-fold increase in GPU power, according to analyst firm Gartner.
There is little doubt that machine learning and pattern recognition-based artificial intelligence is ready to be employed by businesses in a big way. Professional services firm Deloitte Global predicts that over 80 percent of the largest enterprise software companies will integrate AI functionality into their products by the end of the year; by 2020 it expects 95 percent of the top 100 enterprise software companies to have done so.
An important question to ask, then, is how can AI capabilities help businesses, and in which fields will they be most useful?
AI and Routine Processes
According to Tony Williams, principal at legal management consultancy Jomati Consultants, professional services are particularly ripe candidates for the application of this type of artificial intelligence because they rely on large amounts of data and information and “a relatively small amount of judgment.”
What’s key here is that a relatively simple process has to be applied to a large amount of data. Often these tasks are tedious for humans, but artificial intelligences can do them very quickly. There won’t be any human flair or creativity, but the job will be done properly and quickly.
The result is that businesses will be more productive, with AI handling the unglamorous work, ensuring that “processes get applied, stuff is accurate, errors are eliminated, and compliance is met,” according to Dr Stuart Anderson, a research fellow at the Future of Humanity Institute at the University of Oxford.
Something like this has already occurred in the field of journalism, with the Associated Press (AP) employing software that uses artificial intelligence to automatically write earnings reports and minor-league sports stories. AP previously used real journalists to write about 300 earnings reports per quarter, but the smart software now generates 10 times that number of reports, freeing up journalists from what is essentially a data processing task to engage in much higher level and more interesting reporting.
And at Sweden-based furniture company Ikea, an AI helps process employment contracts. This task previously required a human operator to perform 200 clicks in 50 fields to process each contract; the company found it hard to retain staff who could do it because the job was important but dull.
Google is using AI technology in its recently rebranded G Suite office productivity software. For instance, users can employ auto-generated replies for emails that require only a short response. According to Google, 10 percent of replies on mobile devices are sent using this feature, just a year after its release. Another feature, Quick Access on Android, selects files it thinks are most relevant to tasks users are working on, based on interactions with colleagues, recurring meetings and other criteria.
In fact office administration in general is an area where companies are likely to make significant savings by automating processes such as the one illustrated above using software with artificial intelligence.
Future of AI: Chatbots and More
Thanks to the application of machine learning, the capability of software with built-in artificial intelligence can grow so that it encompasses more and more repetitive or dull work. For example, in a report called The Future of White-Collar Work: Sharing Your Cubicle with Robots, Forrester analysts Craig Le Clair and J.P Gownder say that Salesforce.com’s AI team is exploring a chat bot where customers exchange messages with the bot. It can provide information from existing data including FAQs, but when it can’t provide an answer it offers a “get me an expert” link.
“Once the customer hits the link, the chat moves out of the bot’s control and escalates to a human, and the cognitive platform records the exchange with the human. The machine then reprograms itself to take advantage of this training,” they explain.
The more “constrained” a sales scenario, the more likely that a program that uses AI – like a bot – will be able to provide an effective and low-cost alternative to a human. That means that business-to-business sales – which are more constrained that consumer transactions – are most likely to be automated with bots taking on a significant amount of sales and sales support tasks, Le Clair and Gownder believe.
In September Salesforce rolled out Einstein, a suite of artificial intelligence capabilities, to its platform. Among the capabilities it offers to sales people: It can pre-populate case fields and suggest responses, predict customer engagement and deliver emails during optimal engagement windows, and recommend products to shoppers.
“With Salesforce’s launch of Einstein, AI is no longer a novelty but poised to become a key part of mainstream business work flows,” said Scott Horn, CMO at 7, an AI-powered customer engagement solutions provider. “AI promises to fundamentally transform entire businesses and industries. This technology will be commonplace in the next five years.”
Artificial Intelligence in Every Industry
Very quickly AIs are likely to be applied much more widely to areas other than sales and marketing, believes Tom Austin, a Gartner vice president. “This will hit almost every industry,” he said. “We could be talking about professional agents, robots and autonomous transport.”
Likely applications include security and fraud detection (which is already happening with companies like startup Sift Science), AI embedded into human resources and recruiting apps, customer support and retailing, he said.
Pattern recognition is valuable for facial recognition, Austin said, and he foresees a time when returning customers entering a store are recognized by a retail system which then prompts an assistant to greet them and ask if they would like a particular accessory to go with an item they purchased previously.
Artificial Intelligence’s Impact on Human Jobs
While artificial intelligence has the power to transform the way that many companies do business, it’s likely that there will be an impact on the human workforce: Forrester predicts that cognitive technologies will replace 16 percent of U.S. jobs by 2025 and create the equivalent of 9 percent. That means a net 7 percent fewer jobs.
The good news, though, is that man/machine collaboration will mean that artificial intelligence will do much of the drudge work, leaving humans to do higher-level, higher-value and, above all, more interesting tasks.
Paul Rubens has been covering enterprise technology for over 20 years. In that time he has written for leading UK and international publications including The Economist, The Times, Financial Times, the BBC, Computing and ServerWatch.
Paul Ferrill has been writing for over 15 years about computers and network technology. He holds a BS in Electrical Engineering as well as a MS in Electrical Engineering. He is a regular contributor to the computer trade press. He has a specialization in complex data analysis and storage. He has written hundreds of articles and two books for various outlets over the years. His articles have appeared in Enterprise Apps Today and InfoWorld, Network World, PC Magazine, Forbes, and many other publications.