© 2019 SourceMedia. All rights reserved.

Embracing machine intelligence in the enterprise

Artificial intelligence has been part of popular culture for as long as most people reading this have been alive—but it’s increasingly becoming a part of our actual culture too.

We’re not talking about machines that can defeat humans in yet another game or even machines that help search engines to respond to the query “cat” with an appropriate set of feline images—though both of those things are impressive accomplishments. Rather, we’re talking about transformative change, the moment a technology tips from research and proofs-of-concept into widespread use.

Machine intelligence can now identify certain types of cancer as well or more reliably than human doctors, for example. These new machine skills include identifying some cancers’ mutations from mere images—a time-saving feat that humans could never achieve and that could help ensure treatments better match patients’ conditions.

Other use cases emerge almost by the day. Machine learning is being harnessed to help reduce illegal ivory trading. It lets our devices understand us when we talk to them. It’s even starting to screen our phone calls for spammers. And these kinds of advances—of which these examples are only a few—are almost certainly just the beginning.

artificial intelligence in enterprise.jpg
Gouri Mulaka, a senior lab systems engineer for Symantec Corp., walks past patch panels at the company's headquarters in Mountain View, California, U.S., on Tuesday, Aug. 24, 2010. Intel Corp.' $7.68 billion purchase of McAfee Inc. may put pressure on rival Symantec Corp., the largest supplier of security software, to build hacker-thwarting technology inside corporate computers and forge new alliances to stay competitive. Sales will reach $16.5 billion this year in the global security software market according to Gartner Inc. Photographer: Bloomberg/Bloomberg

Indeed, artificial intelligence may eventually impact all industries, which means that if businesses aren’t investigating how this emerging class of technologies can help them now, they may already be at risk of falling behind.

Third-party APIs can help companies get started now

Most enterprises have something they’re world class at—and for all but a few of them, that something is not machine learning. This doesn’t mean that most companies need to spend millions in R&D before they can get started with machine intelligence; it means they may be able to get started by leveraging the work of companies that have already spent millions (or billions) in R&D.

Voice interfaces are a machine learning-driven technology that’s currently hot, for example. If a business senses that its customers are starting to expect voice-enabled services, should it spend years and a small fortune developing the capabilities in-house? Or should it get started now by building its products atop voice technologies already made available by other companies via application programming interfaces (APIs)?

The former option involves cultivating an entirely new, enormously complicated and expensive expertise, and the latter involves using someone else’s expertise to improve whatever the company is already world class at. At many enterprises, it should be a straightforward decision.

In this way, enterprises can begin adopting machine learning the same way they adopt many other technologies that are outside their core competencies: by leveraging the strengths of others.

Mapping capabilities, for example, are central to many businesses, from ridesharing services to coffee houses whose apps help customers find the nearest locations. But many of these companies use someone else’s APIs for their mapping backbone, focusing on what these capabilities can do for their business rather than on creating the capabilities themselves.

Similarly, many of today’s most exciting machine intelligence advances involve organizations leveraging others’ APIs to expand what their industries can accomplish.

The medical examples mentioned earlier, for instance, rely on the core machine learning technologies that make image search—such as those “cat” queries—possible. It’s the combination of forces—the ability for machine learning specialists in one field and doctors in another to amplify one another—that creates many of the biggest innovations. This is the power to participate in ecosystems that APIs can enable.

There are many examples of this, from weather data APIs leveraged by hundreds of applications to mapping and geolocation APIs that thousands of developers use to build new, contextually-aware digital experiences. Machine learning joins these examples as a testament to the ways it may be in a business’s best interest to leverage other parties’ APIs to build or complement its own solutions.

None of this is to say that some legacy businesses won’t transform into machine learning powerhouses or that owning core machine learning technologies won’t confer a competitive advantage. Proprietary advantages, if they truly provide differentiation, will remain as worthwhile as ever. But for many organizations, especially in the short term, the advantage isn’t (and may never be) in possessing the machine learning technology so much as harnessing it for better customer experiences and better business outcomes.

Developer talent is more important than ever—and also changing

Developers are the people who turn a business’s investments in technology into experiences for customers. With respect to machine intelligence, this implies a couple things.

First, the developers who currently create new apps and digital experiences are skilled in UX and modern software design principles, such as building software by assembling data and functionality from disparate sources via APIs. Few within this group will also happen to be skilled in data science, neural networks, and other technologies central to machine intelligence.

Many developers may be able to adapt their skills over time as machine learning becomes embedded in more technologies—but to get started today, they need to be able to leverage machine learning capabilities without being experts in machine learning capabilities. This reasserts the importance of third-party APIs, particularly those supported with robust documentation, sample code, customer service, an active community of users, and other resources.

Second, as skills related to machine intelligence become more common in the future, businesses will need to invest more over time in a new breed of developer who understands not only APIs but also data science and machine learning. Whereas many current developers are skilled writing code that obeys static, pre-defined rules, for example, the new breed of developers focuses on coaxing machines toward desired outcomes, sometimes without fully understanding—let alone deterministically programming—how the intelligent machine arrives at those outcomes.

These new developers are still relatively few in number, which is why it is important to empower developers who may not possess experience with intelligent machines. But over time, if a business has not invested in this new skillset, it may be forever confined to someone else’s APIs and business model, making it difficult to truly expand or innovate.

The intelligent future is already arriving

It’s an exciting time in machine intelligence both because the core technologies are evolving rapidly and because APIs are making these technologies relatively democratically available to all kinds of developers, many of whom would not otherwise have access to nor the means to create such cutting-edge services.

All of this means that the use cases that will define machine learning and influence its evolution are currently being defined by proactive participants, and that those who aren’t participating risk becoming subject to the decisions of those that do. The better future enabled by artificial intelligence with which science fiction has teased us for decades may finally be attainable in the real world. Now is the time for businesses to act if they want to play a meaningful role in that better future as it unfolds.

For reprint and licensing requests for this article, click here.