This content is part of the Essential Guide: Special Report: Artificial intelligence apps come of age

Reality check needed to assess AI applications

When assessing the reality behind today's AI technology, businesses need to think about how it can perform in specific tasks rather than hoping for a do-it-all tool.

The market for AI applications is white hot with huge potential, but that potential needs to be tempered by a heavy dose of realism about the capabilities and business value of artificial intelligence technology, according to industry analysts.

"It's sort of captured the imagination of the world in general, but the danger we have with AI is expectations getting too high," said Mike Gualtieri, an analyst with Forrester Research.

From the early days of computing, the story of AI applications has always been one of early excitement, huge hype and inevitable bust. Every decade or so, some advance in computing power has led to speculation that machines capable of replicating some aspect of human thought were right around the corner. But each time the challenges proved too difficult, and the technology was not ready.

What's different this time is cheap storage, which has allowed companies to stash huge troves of data, a critical need for training machine learning algorithms -- the "brains" behind artificial intelligence. At the same time, computing power has increased to the point where algorithms can churn through all this data nearly instantaneously. The growth in cloud computing means that smaller and midsize companies can access high-powered AI tools through subscription models without having to build them themselves. The combination of these factors has led many industry watchers to say, with some important qualifications, that AI might be for real this time.

The AI platform war is on

Rather than selling AI software directly, most of the big companies in this space provide the technology as a platform, managing the machine learning and natural language processing components themselves while allowing customers to build applications on top of the engine. This platform approach is expected to be a major area of competition in the AI market.

Facebook announced this month that it would allow businesses to build chatbots using the AI engine in its Messenger app. Microsoft made a similar announcement last month. IBM has been one of the bigger players in the AI platform space ever since it made Watson available to developers. So far developers have used it to build smarter travel planning assistants, shopping recommendation engines and health coaches.

Jerome Pesenti, IBM's vice president of Watson core technology, said he believes these kinds of smart virtual assistants are going to be everywhere in the next few years and IBM wants Watson to be at the center of this trend.

"I believe that we're at the start of a new era of customer interactions," he said. "It's a change in the paradigm in the way companies interact with customers."

But it's not just the big tech companies that want to grab a piece of the AI platform market. David Schubmehl, a research director at IDC covering cognitive systems, said he's currently tracking 25 to 30 companies who offer some kind of AI platform.

Schubmehl said he thinks there is room for maybe four to five platforms in the market going forward. But right now there's so much development happening that it's not clear which platforms are going to win out or what kind of capabilities customers are going to demand.

"I do think there will be consolidation, but right now we're in the Wild West phase," he said. "That's the next big battle in terms of platforms."

Believe the hype -- at least some of it

Businesses are already starting to embed AI in systems to make them smarter. Insurance companies are looking at applying it to the process of approving medical claims. Retailers are applying it to customer service and marketing. And enterprise technology companies like Salesforce are looking to embed it in their software.

But even as businesses are finding real value in AI applications, there's a widening pitfall. Success breeds hype, which itself leads to inflated expectations. Should burgeoning AI software fail to live up to unrealistic expectations, it could brew disappointment and stain the technology.

"If you took it as a horizontal technology and apply it to everything, that's setting it up for another failure," said Rajeev Ronanki, who leads Deloitte Consulting's cognitive computing practice. 

Some see AI potentially following a similar trajectory as big data, which burst on the scene with seemingly unlimited potential, drew huge hype and then has come to be recognized as important, but less than transformational.

"All of the things we take as big data applications have gone out of the hype cycle and are into everyday use," Schubmehl said. "That's what we'll see happen in the AI market too."

Mixed expectations for AI applications

If AI projects are focused and narrowly tailored to specific tasks, some believe they could soon play an important role in businesses' operations.

"Organizations that are able to harness this technology and use it as a disrupting influence in their markets are going to have a significant competitive advantage," Schubmehl said.

Ronanki compared the situation today to what Netflix did to Blockbuster video rental stores a decade ago. Businesses that embrace AI and use it in smart ways could shift dynamics in their lines of business and gain a significant edge. In terms of changing the way the businesses interact with customers, he said AI could be no less revolutionary than the assembly line was in manufacturing last century.

"There's a clear, compelling business value and the technology is ready," he said.

But Forrester's Gualtieri said many businesses are still in the early days of figuring out how to apply predictive analytics to specific business problems. For these organizations, chasing AI applications would be premature.

He pointed out that even some large tech companies like Google and Microsoft have had mixed results generating revenue and improving business operations by using AI. And a typical medium- to large-sized enterprise can't expect to do better than large tech companies, with all the resources they can apply to the problem.

"The real value for enterprises is to define predictive models," Gualtieri said. "AI is something to research and understand what's happening. But my advice is to focus on predictive models, become a predictive enterprise."  

Next Steps

Recent computing failures say little about true AI technology

AI could come to play a key role in software testing

Even in an AI-dominated future, humans still have a role

Dig Deeper on Enterprise applications of AI