AI Is Moving Into Mainstream Enterprise Production
An IBM Institute for Business Value study recaps how artificial intelligence is moving into mainstream enterprise production.
By Adam Oxford04/01/2019
“Bonjour Djingo, comment transférer 100 euros sur mon compte d’épargne?”
Figuring out how to transfer money to a savings account shouldn’t be too difficult for the French. They’re a sophisticated nation when it comes to banking: According to a study by Next Content, 47 percent of banking customers in the country use a smartphone app to interact with their bank at least once a week (oran.ge/2De1hva ).
That’s one of the reasons why, when telecommunications giant Orange decided to launch its own bank in 2017, it leaped straight ahead and embedded the next-generation of customer service into its products from the start.
If you’re an Orange customer, chances are that you’ve chatted to Djingo, a virtual advisor powered by IBM Watson* that’s capable of answering natural language questions and executing requests such as canceling cards, checking payments and more.
Djingo works 24-7. At least a fifth of the 90,000 queries it answers per month come outside of normal office hours. Like other virtual assistants, Djingo can hand over communication to a human operator if a question proves too tough, but it isn’t in the habit of doing that too often.
“Djingo is the customers’ first point of contact,” explains Djamel Mostefa, head of artificial intelligence (AI) at Orange Bank. “It provides customers with all of the information they need at all times, wherever they may be. It understands around 80 percent of the questions asked in a natural language and can even carry out some banking operations.
"Since the launch of Orange Bank, more than 1 million conversations have been initiated by our clients with Djingo, 50 percent of which are handled entirely by the virtual advisor without any redirection to the Customer Relationship Center.”
What’s more, at the end of 2018, Djingo expanded beyond banking services and into the rest of the group’s consumer-facing activities. Orange has shown off a Djingo-powered smart speaker comparable to Amazon’s Alexa, Google Home or Apple’s Homepod. The Djingo speaker will be able to control connected devices around the house, place calls, send texts and—of course—continue to grow its capabilities as a banker by learning more about Orange’s customers.
“In the coming years, Djingo will evolve at Orange Bank to become a full virtual financial coach, capable of anticipating customers’ expenses based on spending habits or recommend savings for a specific project, and proactively suggest services adapted to their needs,” says Mostefa, “It will even be able to understand emotions such as stress through a keyword recognition system.”
“You see all the headlines about the robots replacing jobs. But what we saw was contrary to that. The strategic rationale is all about growth metrics.”—Brian Goehring, Associate Partner, AI / Cognitive & Analytics, IBM Institute for Business Value
Accelerating AI Implementation
Djingo is just one example out of many in which AI-driven initiatives are maturing around the world, and proving the business case for the technology. According to statistics from Gartner’s 2019 CIO Survey (gtnr.it/2RRE4ax), the use of AI is accelerating at an incredible rate: Some 37 percent of CIOs queried say that their firm has implemented at least one AI project.
IBM’s Brian Goehring is keenly aware of the rate of change in AI deployments, and the challenges that businesses face today. As an associate partner and the AI/cognitive and analytics lead at the IBM Institute for Business Value (IBV), he’s a key voice in helping shape strategy in the field and a co-author of the recent IBV report, “Shifting Towards Enterprise-Grade AI: Resolving Data and Skills Gaps to Realize Value” (ibm.co/2WVh6Sm).
That report, Goehring explains, is based on interviews with 5,000 C-suite executives, heads of department across 13 different business functions, including IT, human resources, marketing, sales and finance. It was based on a similar study conducted in 2016, giving Goehring and his co-authors the ability to identify the way that AI deployment has changed.
“What we saw two years ago was that they were really throwing spaghetti at the wall and looking to see what sticks.” Goehring says. “Companies just didn’t know where AI would be most valuable to them.”
In 2016, the report authors found that 47 percent of executives surveyed still believed that AI was more hype than value. That has changed, and an additional one third of organizations surveyed are implementing AI today.
“What we see is companies moving beyond proof of concept stage and ‘AI tourism,’ ” he says, “They have a much more pragmatic and practical take on the subject. They’ve moving beyond dabbling with AI to implementing AI systematically across the enterprise.”
Real-World Use Cases
Widespread implementation of AI is a long way off yet, but very specific use cases are emerging as the forerunners for AI trials. In 2016, IBV asked CEOs what business functions they thought would benefit from AI, and each of the 13 business areas studied were selected by at least 67 percent of respondents. AI, it seemed, could be anywhere. In 2018, that had changed dramatically, and just IT, information security, innovation, customer service and risk scored above 50 percent.
“This time around, they’ve really honed in on five specific areas, and are a lot more discerning about what they perceive as value,” says Goehring. “That’s not to say that in HR and procurement there isn’t value in AI. It’s a reflection of the CEO prioritizing where it can deliver the most value.”
Djingo is in good company. Other organizations to deploy customer support bots using Watson include Macy’s, Staples and CAD specialist Autodesk. Watson technology has also been used to detect fraudulent financial transactions based on customer data, to help flight online ads more effectively and as a diagnostic assistant for healthcare providers (see "Doctor, Doctor"). It’s even used to guide prospective IBMers to career opportunities within the company.
What may surprise some, Goehring explains, is that business leaders appear to have a different perception of the value of AI compared to the popular press and opinion. “You see all the headlines about the robots replacing jobs,” he says. “But what we saw was contrary to that. The strategic rationale is all about growth metrics.”
To put it another way, AI isn’t being used to downsize staff numbers, but to increase sales figures.
This is reflected in the report in which respondents answered questions about what is driving AI adoption. Overwhelmingly, these biggest drivers are customer satisfaction and customer retention, followed by reducing the cost of new customer acquisition—arguably also a growth metric.
According to Goehring, operational cost reduction is important. “CFOs still have to approve budgets, so it’s essential that the business case flies. But we certainly see that top line numbers are the most important aspect,” he says.
Critically, even though spending decisions have rationalized on a few business activities, spending levels themselves are rising—and most firms expect that they’ll continue to increase budgets for AI over the next three years.
“When you look at training AI systems and learning through ingesting data, that's not something that can be replicated or bought off the shelf. Incumbents have an advantage as they have lots of data to leverage—customer data, financial data, logistics data and more. It's a competitive and comparative advantage at a company, industry and country level.”—Brian Goehring
Opportunities and Roadblocks
While AI use has increased dramatically, Goehring is quick to point out that it’s still not exactly mainstream yet, and still many more use cases are yet to be discovered.
“I don’t want to overstate the case,” he says. “We are beyond the peak of the hype cycle, but we have yet to see a large number of companies who are fully operationalizing AI,” he explains.
What is a concern, Goehring notes, is that the uptake of AI isn’t evenly distributed, and while it makes a certain amount of sense for organizations that aren’t heavily digitized or well off to let those better able to take advantage of AI bear the brunt of research and development today, catching up tomorrow could be challenging.
“When you look at training AI systems and learning through ingesting data,” Goehring says, “that’s not something that can be replicated or bought off the shelf. Incumbents have an advantage as they have lots of data to leverage—customer data, financial data, logistics data and more. It’s a competitive and comparative advantage at a company, industry and country level.”
For the latter, Goehring says, a definite geographical pattern exists to the uptake of AI, and emerging economies are in danger of getting left behind, with a huge gap between the AI haves and have-nots.
Ultimately, much of the gap is likely to come down to skills. According to Gartner, technology is no longer the limiting factor in AI deployment. Rather, the biggest struggle is recruiting people to design and implement strategies. Fifty-four percent of respondents to the Gartner CIO survey said that having the correct skills is the biggest challenge they face, while 63 percent cited it as the biggest barrier to AI success in the IBV study.
For implementations to be successful, Goehring adds, a wider discussion is needed around what skills are necessary. Certainly, a shortage of data scientists and high-end application designers exists, but companies also need to think about soft skills.
“Companies are struggling with the ‘comfort level’ around AI,” he says, “Customer agents don’t need to know about Python and feedback loops, but they do require a level of familiarity with AI so that they can ensure a seamless journey for the customer. They need to be able to see what it was in the interaction with the chatbot that wasn’t successful, for example, so that they aren’t repeating questions.”
Having AI co-workers, in other words, will indeed be a culture shock for many—just not in the way that the popular depictions might have it. As the authors of the IBV report put it, however, businesses must manage all of these cultural and technical changes and be prepared to adapt quickly. Otherwise they may miss the opportunity to realize the full potential of enterprise-grade AI.
Learn more about how AI is moving into enterprise production by reading the IBV report: ibm.co/2WVh6Sm.
According to Frost and Sullivan, the market for artificial intelligence (AI) in healthcare is expected to grow by 68 percent year-over-year between now and 2022. Its application is expected to be broad, from supply chain management to helping physicians develop personalized care plans based on patient history and the latest research.
One firm making headlines at this January’s Consumer Electronics Show (CES) in Las Vegas was IBM Watson partner HeartBit. It launched a range of T-shirts with built-in electrocardiograph (ECG) monitors, capable of reading heart muscle behavior similar to the machines common in hospitals.
The first generation of clothing and applications from HeartBit are designed to help fitness fans improve their workout regime, and data from the ECG system is analyzed using AI to recommend new training programs. In the future, however, HeartBit plans to release a version that will help treat patients recovering from heart attacks.
Adam Oxford is a freelance writer based in South Africa. He’s covered technology-related issues for more than 20 years.
Post a Comment
Note: Comments are moderated and will not appear until approvedcomments powered by Disqus