One of the cornerstones of International Business Machines' (IBM 0.25%) ongoing transformation is cognitive computing, which encompasses artificial intelligence and other related technologies. IBM is a business that serves other businesses, and its approach to AI stays true to its purpose. IBM Watson, the company's well-known AI system, is being used in industries like healthcare and financial services to augment the skills of professionals in those fields. The long-term potential of the technology is immense.
IBM has made a bet that cognitive computing will be a big part of its future. Here's how the company got here, and how it plans to win.
A (very) brief history of IBM
IBM is a rare example of a tech company that has been able to adapt and evolve, remaining relevant through many decades and technological eras. Since its founding in 1911 as the Computing-Tabulating-Recording Company, IBM has gone through multiple, sometimes painful, transformations that tested the company's resolve.
In its early days, IBM sold and leased mechanical tabulating machines, as well as typewriters and other devices. The company's first major transformation came at the dawn of the computer age. IBM introduced its first large computer based on vacuum tubes in 1952. Computers based on transistors came later that decade, marking the birth of IBM's mainframe business. IBM went on to dominate the personal computer market as well, at least for a while. The company's mechanical tabulating machine business was a cash cow, but IBM wasn't afraid to disrupt itself when it became clear that computers were the future.
IBM's second major transformation came in the 1990s. After decades of success, complacency led the company to nearly implode as the core mainframe business struggled and the PC business became increasingly commoditized. IBM had become bloated, and it lost sight of where the industry was going. An outsider CEO was brought in, saving the company by slashing costs and betting on services. IBM became an integrated provider of solutions spanning hardware, software, and services, a model that then drove its growth for the better part of two decades.
That brings us to IBM's current transformation. Over the past five years or so, IBM has undergone yet another reinvention, this time betting on cloud and cognitive computing to fuel its business in the years to come. Legacy businesses have been sold off or de-emphasized, while resources have been poured into growth areas like artificial intelligence. The company has built a cloud computing business with nearly $20 billion of annual revenue, and its Watson AI system has been applied to a wide variety of industries and applications.
Watson is the face of IBM's effort to cement its status as a leader in the fast-growing AI market.
The birth of IBM Watson
The world was introduced to Watson in 2011 when IBM put the AI system up against two champions on the game show Jeopardy! Watson cleaned house with an impressive but imperfect performance, able to parse the clues and provide correct responses often enough to win the game and represent a major AI breakthrough.
At the time, Watson was powered by 750 of IBM's Power servers, featuring a total of 16 terabytes of memory and 2,880 POWER7 CPU cores. Today, Watson-based services are available via IBM's cloud, requiring far fewer resources than the initial version. By 2013, IBM had managed to nearly triple performance while reducing hardware requirements to a single Power server.
One important thing to know: Watson is not a one-size-fits-all system. The original Watson system used on Jeopardy! was trained specifically for that purpose. A different application requires training with different data. For example, IBM offers a visual-recognition cloud service powered by Watson; this version of Watson is trained with images tagged with the objects they contain, and the result is a system that can identify objects in other images. You can view Watson as a set of AI technologies that can be applied to different tasks, rather than a monolithic AI system.
Much more than machine learning
The term artificial intelligence is often used interchangeably with two other terms, machine learning and deep learning. But that's not quite accurate. Machine learning is a set of algorithms that improve through experience. A machine-learning system that identifies objects in images, for example, gets better as it's trained on more data.
Deep learning is a subset of machine learning, providing a framework that can lead to improved results compared to other algorithms. An artificial neural network is one type of deep-learning architecture, based roughly on how the human brain works. Many of the advances in AI we've seen over the past few years, like self-driving cars, have been enabled by advances in deep learning.
Artificial intelligence is much broader, encompassing any algorithm that can mimic the cognitive abilities of humans. This includes natural-language processing, planning, perception of the environment, and machine and deep learning algorithms.
A wide variety of AI technologies sit under the Watson umbrella, including machine learning, deep learning, voice recognition, sentiment analysis, and natural-language processing, to name a few. The original version of Watson was a question-answering system. Today, Watson offers a broader set of capabilities, and is able to ingest massive quantities of structured and unstructured data.
IBM's AI strategy
IBM is not a consumer-facing company, so you won't be seeing any gadgets coming from IBM powered by Watson. Instead, IBM's focus is on providing solutions to its enterprise customers. The company's 2017 annual report says:
IBM Watson's enterprise-strength artificial intelligence is transforming the way people work in nearly every industry. It helps organizations derive insight from complex and unstructured information. And it allows professionals to scale their expertise and focus their efforts on higher-value work.
While other large technology companies are aiming their AI efforts at smart speakers and self-driving cars, IBM is working to provide AI solutions to the wide range of industries and organizations that it serves. IBM counts as customers 97% of the world's largest banks, 80% of global retailers, and 83% of the world's largest communications service providers. As companies of every stripe look to invest in AI, IBM wants to be the first and best choice for enterprise organizations.
IBM sees AI as a way to augment human intelligence, not replace it. In areas like healthcare and financial services, where Watson has been deployed for various applications, the system acts as a tool for professionals, allowing them to do their job better.
Thomas Watson, Jr. was CEO of IBM from 1956 to 1971, during the period when the company bet on computers. One of his quotes sums up IBM's view of AI: "Our machines should be nothing more than tools for extending the powers of the human beings who use them."
One way IBM attempts to differentiate its AI offerings from the competition is data ownership. IBM wants to be the most trusted name in AI. As it said in the 2017 annual report:
We also believe that data and the insights it generates belong to their creators. No one should have to give up ownership or control of their data to benefit from AI and cloud computing. We have built and are deploying Watson accordingly.
Here's IBM's AI strategy, summed up in three points:
- Focus on serving enterprise customers.
- Apply AI to areas where it can augment human intelligence, increasing efficiency or lowering costs.
- Give clients control and ownership of their data and insights.
AI is still an early-innings technology, but IBM has already put it to work in a wide variety of industries.
Total healthcare spending in the United States reached $3.5 trillion in 2017, and the U.S. Centers for Medicare and Medicaid Services projects spending to reach $5.7 trillion by 2026. If there was ever an industry that could benefit from increased efficiencies and lower costs enabled by AI, healthcare is it.
IBM views Watson Health, a business unit aimed at applying Watson AI technology to the healthcare industry, as the company's "moonshot." It's an ambitious undertaking that will be slow to generate much in the way of profits, but has almost limitless potential. Watson Health currently touches well over 100,000 patients and consumers, and that number could rise exponentially in the coming years as AI makes further inroads in the healthcare industry.
What exactly is Watson up to in healthcare? A lot, as it turns out. Here are a few examples of how Watson is changing the healthcare industry:
IBM Watson Care Manager is used to create individualized care plans and recommend optimal approaches; it processes vast amounts of structured and unstructured data, including evidence-based research, quality standards, and regulatory requirements. Roughly 20% of patients consume 80% of total healthcare costs, according to IBM Watson Health general manager Deborah DiSanzo. Deploying Watson for those patients can lower costs while producing better outcomes.
Around 147,000 patients now have their care plans managed with the assistance of IBM Watson Care Manager, with the AI system proving to be most useful for behavioral health and social care management.
IBM Watson for Drug Discovery is aimed at accelerating the process of bringing new drugs to market, which takes ten years on average for biotech and pharmaceutical companies. IBM announced a collaboration with Pfizer in late 2016 to use Watson for Drug Discovery to identify new drug targets, combination therapies for study, and patient selection strategies in immuno-oncology.
In another partnership, IBM and Barrow Neurological Institute tapped Watson for Drug Discovery to explore unidentified genes and proteins that may be linked to amyotrophic lateral sclerosis. The study found five genes associated with ALS that had never been linked before, and did so within a few months; it would have taken years without Watson, said Barrow scientists.
IBM Watson for Oncology is used to provide individualized, evidence-based cancer care for patients based on millions of data points, with the goal of improving the consistency and quality of cancer care. Watson acts as a tool for doctors, one that's able to consider a vast amount of data and information and make recommendations for treatment. The system has been deployed in 155 hospitals and healthcare organizations, and it's been trained on a variety of cancer types.
IBM Watson for Clinical Trial Matching aims to reduce the time it takes for clinicians to find clinical trial options for patients, by eliminating the need to manually compare enrollment criteria with patient medical data. The AI system reduced prescreening wait times by 78% during a 16-week trial period, and it automatically eliminated 94% of patients who were ineligible for clinical trials, according to DiSanzo.
Watson Health, and AI in general, are not silver bullets for the healthcare industry. Watson won't be replacing doctors. Instead, Watson Health is a collection of tools for healthcare professionals that can make processes more efficient, lower costs, and potentially improve patient outcomes.
IBM made a long-term bet when it thrust Watson into the healthcare industry. Watson Health likely isn't contributing much to the company's financial results right now. But that could change in the coming years and decades.
IBM announced the acquisition of Promontory Financial Group in late 2016, betting that Watson could be useful churning through the mountain of regulations related to the financial services industry. Promontory is a global risk management and regulatory compliance consulting firm that employed 600 professionals at the time of the acquisition. The plan was to have Promontory's highly trained professionals train Watson on regulatory information, eventually using the AI system in real-world applications.
Regulatory compliance for the financial services industry is a near-perfect application for AI. The amount of data is vast. More than 20,000 new regulatory requirements were created in 2015 alone, and IBM projects that the total catalog of regulations will surpass 300 million pages by 2020. More than 10% of operational spending at major banks is related to regulations and compliance, totaling $270 billion annually, according to IBM. Watson doesn't need to move mountains to have a major impact: Even shaving off a few percentage points of the cost would add up to billions of dollars in savings for financial firms.
Watson is also being used in tax preparation. H&R Block began using Watson last tax season to help its tax professionals make sure that clients receive every possible tax deduction. The U.S. federal tax code contained 74,000 pages as of early 2017, and thousands of changes are made every year. The tax reform bill passed in late 2017 only makes things more complicated.
Massive amounts of data and inefficient processes make financial services a tailor-made industry for IBM Watson.
Beyond healthcare and financial services, IBM Watson is being used for a wide variety of other applications. The number of organizations using Watson grew by more than 70% in 2017, and that growth should continue as the adoption of AI expands.
IBM Watson is being used:
- By pipeline engineers to anticipate and prevent pipeline failures: Watson is capable of predicting failures up to six days in advance, a 24-fold improvement.
- By General Motors for its OnStar Go system: This Watson-powered system will learn drivers' preferences and deliver individualized location-based interactions.
- By lawyers to sift through more than one million legal documents per second and find relevant information for cases: New associates spend one-third of their time looking at laws, prior cases, and legal journals, according to a study cited by IBM.
- By workplace messaging company Slack to power its customer-service bot. Watson services are also available to developers building bots and other tools for the platform.
- By companies looking to build industry-specific, branded, and customized virtual assistants: IBM announced Watson Assistant earlier this year, powered by various cloud-based APIs. Companies that build applications based on Watson Assistant own and control the data.
- For streamlining customer support: For example, IBM won a $24 million contract from New York City in 2017 to overhaul its 311 system with a new system powered by Watson.
These are just a few of the use cases for IBM's AI technology. IBM CEO Ginni Rometty said in 2016 that the company expects the number of consumers touched by Watson, either directly or indirectly, to reach more than 1 billion by the end of 2017. Despite that number, IBM is still at the beginning of its AI journey.
A silver thread
IBM doesn't disclose how much revenue Watson generates, so it's hard to estimate the financial impact. Back in 2013, Rometty reportedly set an ambitious goal for Watson: $10 billion of revenue within ten years, with $1 billion of revenue by 2018. Whether IBM is on track is unclear.
One reason IBM doesn't disclose Watson-related revenue is that Watson is involved in so many of IBM's products and services. Executives have described Watson in the past as a "silver thread" that runs through IBM, meaning that it touches a lot of different areas. Attributing revenue to Watson, when it's the products and services that integrate Watson generating that revenue, may not make all that much sense. Still, given that Watson is the face of IBM's transformation, the lack of disclosure suggests that the company may have fallen behind on achieving its goal.
Because Watson is an AI system, employing machine learning and other technologies that require extensive training with data, it can take a long time for the system to be useful for a particular application. You'd expect revenue generation to be very slow or nonexistent at first, then increase rapidly if the system proves capable. Even if Watson isn't generating all that much revenue today, revenue could rise quickly as the technology is adopted.
The greatest opportunity of our time
During her speech at the Think 2018 conference, Rometty discussed how important AI technology will be:
It is a moment that happens every 25 years. It's something that happens when business and technology architectures are changing at the same time. It has the potential to change everything.
She added: "AI is the greatest opportunity of our time."
The potential for AI to change and disrupt industries is incalculable. Inefficient processes can be streamlined, and costs can be cut. Brand-new business models enabled by AI systems can emerge. In the same way that the internet and the proliferation of mobile devices generated geysers of innovation, so too will AI.
IBM wants to be the AI company for large enterprises. The company's vast base of existing customers, all dependent on IBM technology, gives it a significant competitive advantage. And its insistence that data and insights belong to the customer differentiates it from the competition.
It's hard to pick winners and losers in AI, because the story of the technology will play out over many years. But IBM is well-positioned to reap the benefits of the AI revolution.