Computers soon could deliver smarter healthcare to patients at Massachusetts General Hospital (MGH).
That’s the goal of a new partnership between Partners HealthCare, the hospital’s parent organization, and GE Healthcare. The two Boston-based institutions in May announced a 10-year collaboration to develop and integrate artificial intelligence throughout Partners’ clinical operations.
Partners hopes A.I. can improve patient outcomes and increase clinician productivity. The nonprofit healthcare system plans to first use A.I. to enhance diagnostic imaging, with intelligent systems being developed to detect, for example, even minute changes in tumors and then use data analysis to determine optimal treatments tailored to each case.
This is the future of medicine, but it’s been a long time in the making. The medical community has been laying the groundwork for such advances for years, says Dr. Mark Michalski, executive director of the Center for Clinical Data Science, a collaboration of MGH and Brigham and Women’s Hospital.
“We’ve been collecting and annotating data for decades without knowing that this was when artificial intelligence was going to take,” Michalski says. “The tools are now quite good, and the potential is astounding.”
Partners officials say they expect to create new business models for applying AI to clinic work, with roles for the technology from patient admission to discharge. Early applications could include determining the prognostic impact of stroke, identifying fractures in emergency room patients and indicating the likelihood of cancer on ultrasound.
A.I. is the next wave of business technology innovation, and organizations such as Partners are charging ahead with initiatives to identify and develop the use cases that will yield the best early results. But there’s plenty of work to do before even getting to that stage, with the bulk of that work falling to CIOs and their IT departments. Now is the time to prepare, analysts say.
“This is top of mind for CIOs and even CEOs. And although last year it was being researched more than it was actually being used, we are seeing more use it or have concrete plans to use it,” says Forrester Research analyst Mike Gualtieri.
A Forrester report called “Artificial Intelligence: What’s Possible for Enterprises in 2017” found that only 12% of the 391 business and technology professionals it polled are currently using A.I. systems. However, 58% are researching A.I. technologies and what it takes to support their use in the enterprise, and 39% are identifying and designing A.I. capacities to deploy. The report, published in November 2016, also found that 36% of respondents are educating the business or building the business case about A.I.’s potential.
The time for A.I. is now
The exuberance around A.I. technologies right now lends an exaggerated sense of newness to the field. The concept and the building blocks of A.I. are, in fact, decades old. And as the Forrester report points out, pure A.I. — computers that mimic or even exceed human intelligence — is still not reality.
Yet analysts and organizational executives say they’re increasingly harnessing A.I. and A.I.-related technologies (sometimes called pragmatic A.I.), such as machine learning, image analysis, deep learning, robotics and speech recognition, to gain competitive advantages and realize returns on investments.
“Everyone is trying to figure out how to get into the A.I. game,” says Matthew Lieberman, advisory marketing leader at PricewaterhouseCoopers (PwC).
There are, to be sure, plenty of challenges to reaching those goals.
“Companies still don’t know how they should use A.I. and what the benefits will be and what they should spend on it,” Lieberman says, adding that having a lot of data — the fuel for A.I. programs — doesn’t mean an organization is ready for the leap to A.I. “You have to first ask what you want to do with it.”
A.I. does have significant benefits, Lieberman says. It can reduce human error, inform business strategy, improve growth and reduce administrative tasks — all of which can produce substantial ROI.
Lieberman points to a recent PwC survey of 2,500 U.S. business and consumer leaders that showed that 72% of them believe A.I. will be the biggest business advantage of the future. The report, “Bot.Me: A Revolutionary Partnership,” released in April, also found that A.I. further shifts human tasks from menial to strategic. Given such confidence in the technologies, Lieberman says, investments in A.I. will be the norm in three to five years.
Meanwhile, Tractica, a market intelligence firm that focuses on human interaction with technology, forecasts that the revenue from enterprise A.I. applications will increase from US$358 million in 2016 to $31.2 billion by 2025.
PwC’s Lieberman points to a few current high-profile uses of A.I., such as self-driving cars but notes that there are more examples of organizations using A.I. technologies behind the scenes — for example, marketers’ use of these technologies for personalized advertising campaigns and individualized delivery of content.
Still, he says, the use of A.I. in corporate America remains limited.
“Those who are employing intelligent systems are still low. Those who are testing and exploring goes up higher. And those who are interested in it and see it as a key differentiator are the majority,” he says.
Foundational elements are critical
LinkedIn, the professional social site, is one company that has figured out what it wants to do with A.I. and is already deploying it. The 15-year-old company started investing in A.I. technologies in 2007, spurred on by its belief that its data “is our biggest asset,” says vice president of A.I. and machine learning Deepak Agarwal.
Agarwal says machine learning powers many LinkedIn features, such as ranking search results for users and selecting which advertisements, news feed updates and recommended connections to display for each member. “It’s like oxygen for our product,” he says.
A.I. technologies are like every other modern technology, Agarwal says, in that their success owes a lot to other, foundational elements such as, in the case of A.I., automation and adequate compute power.
“I don’t believe that A.I. today is a turnkey solution. It’s gotten better, but by no means is it a turnkey solution,” he adds.
Agarwal says his company first focused on building a robust data management system and creating structured data sets that could yield the kinds of insights and information needed for A.I. technologies to produce results.
Furthermore, he says, LinkedIn’s leaders understood early on that identifying the right uses of machine learning and A.I. was critical for success as well. He notes that part of the A.I. strategy was, and remains, as much a business issue as it is a technology exercise.
“It always starts with the right question. … There’s no point to using artificial intelligence to solve the wrong problem,” he says. More specifically, Agarwal says, LinkedIn sees A.I. as enabling the company to “make sure that everyone on our platform gets the right career opportunity.”
LinkedIn ensures that its A.I. efforts focus on the right questions by having members of Agarwal’s 250-person team work directly with the various divisions, such as the product group, he says, and by looking for employees who understand the complexity of the company’s business needs as well as what the technology can do and what it needs to do.
“We need technologists who have a good sense of the business,” he says.
Farmers Insurance, an 89-year-old company headquartered in Los Angeles, is also moving forward with its A.I. endeavors. CIO Ron Guerrier says his company is using the Salesforce Intelligent Customer Success Platform to transform how its employees and agents deliver services, with an aim to help them meet changing customer expectations.
“We will leverage A.I. to enhance the abilities of the employee, agent and customer and make us smarter at what we do,” he says, stressing that Farmers Insurance sees A.I. as a tool that can help build strong relationships between the company’s representatives and its customers. A.I. should provide employees with specific, customized information about each customer they’re helping, so the company can best address each one’s unique needs and requirements, he says.
Delivering on that goal is challenging, with a strong data program being only one prerequisite. Echoing Agarwal, Guerrier says that other essential elements are building the technology infrastructure that can support the company’s data, analytics and digital efforts and having the right talent and skills in place to make sure that infrastructure runs as it should. For example, the company has a chief data officer, a digital officer and an executive overseeing the IT piping who ensure that their individual areas are solid in their ability to support the company’s A.I. efforts.
IT must also focus closely on the technology components that integrate A.I. into its systems, some of which are legacy applications 20 or more years old, Guerrier adds. “You can’t just dismiss that. You have to focus on that, because those are brittle. Those can break,” he says.
Additionally, the CIO and IT must be spot on when it comes to orchestrating all these components, to ensure that all the pieces not only work but work together, he adds. That must also include wrapping in the business use cases.
“The collaboration with the business is absolutely critical. If you’re doing this in a bubble, building it in a way that doesn’t add value, then you won’t get the funding,” Guerrier says.
Most organizations aren’t as far along as LinkedIn and Farmers. Instead, many are still struggling with maturing the technologies and related procedures upon which their A.I. layers will rest, says Aditya Kaul, a research director at Tractica.
“Most companies and enterprises are still going through their digital transformations; Some of them are further ahead, some are still trying to come to grips with it, and some are in the middle of it,” Kaul says. “Then comes A.I., which is the next leap.”
Marketing and advertising are already recognized leaders in their use of A.I. technologies, employing them to determine what merchandise to tempt consumers to buy by parsing copious amounts of data that no team of humans could possibly analyze, much less as quickly as machines can.
Some industries are using A.I. to create efficiencies. Kaul points to some law firms that are using A.I. technologies for contract analysis, completing in minutes tasks that would take junior associates all day.
Other emerging use cases show how A.I. can be harnessed to do jobs that would have been impossible otherwise, Kaul says. He cites the public-safety use of A.I. to read drivers’ faces as they cruise along to determine if they’re sleepy, distracted or otherwise unsuited to drive and then alert them that they’re unsafe behind the wheel.
Building up A.I. capabilities
Like Agarwal, Kaul says these and other successful A.I. use cases need robust data programs, new expertise and cutting-edge software and hardware. Those are big investments for most companies, so he says executives ease into A.I. by building pieces of their intelligence infrastructure, starting by hiring the data scientists, analytics professionals and experts in machine learning and deep learning who can develop and train the right algorithms.
Companies then must either build or buy their intelligent platforms, Kaul says. And they need to acquire the right compute power, whether through on-site, high-performance systems or GPU clusters in the cloud.
Given those high-ticket investments, the time frame for A.I. enterprise adoption is quite long, Kaul says.
But he says some organizations are already thinking long term and “actually see A.I. as a way to transform.” For example, the insurance industry is considering scenarios such as customers using their smartphones to upload photos of their damaged cars. Intelligent systems would then assess the damage and offer settlement estimates — saving the customers the inconvenience of having to go to adjusters.
In many companies, analysts say, delivering on such visions will likely fall to emerging roles with titles such as head of A.I. or head of intelligence, or possibly to the innovation group.
Rest assured, though, that the CIO and the IT department will remain critical for the success of A.I. within an organization, since they’re responsible for most of the major components — from the computing resources to the data operations — that power intelligent initiatives.
But IT leaders must be onboard if they don’t want to be left behind, experts say.
“The biggest problem with CIOs in this whole A.I. space is that a lot of CIOs have been around for a while. And they may be overly dismissive of it, or they might not embrace it quickly enough. There are many who are skeptical, and that skepticism can come out and lead to inaction,” Forrester’s Gualtieri says.
And to be clear: Companies whose IT departments don’t guide A.I. strategies or can’t deliver the capabilities needed to support A.I.-related programs will find themselves left behind, analysts say.
“One of the core applications of A.I. is prediction: predicting outcomes, predicting what someone might buy or when a machine will break down,” Gualtieri says. “And when you can predict something, you have an advantage.”