AI in IT tools promises better, faster, stronger ops
A comprehensive collection of articles, videos and more, hand-picked by our editors
With another tsunami of hype about the wonders artificial intelligence can bring, IT professionals would be wise to seek higher ground until they can sort out exactly how machine learning business applications can help them.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
We have seen this hype a couple of times before. On and off through the mid-1980s and 1990s, a variety of AI-flavored prototype technologies promised to automate out of existence any number of technology issues bogging down IT productivity.
But each time, the hype misled many potential customers who were disappointed in the commercial products -- if they were delivered at all -- and the air quickly came out of the AI balloon. Visions of American businesses resembling something out of Star Wars ended up star-crossed.
"In the 1980s, there was an enormous amount of AI startups that collapsed after a few years because, in part, expectations were too high," said Mike Gualtieri, vice president and principal analyst with Forrester Research. "People still associate AI with sci-fi movies or something like Westworld instead of something more practical," he said.
But with significant advances in software and hardware, particularly over the past decade, along with the delivery of AI-laced products consumers use every day -- such as Amazon Echo, Google Home and Microsoft Cortana -- it is unlikely the AI market will repeat the failure of the 1980s and 1990s. Still, IT shops will have to make prudent choices as they take their first steps embracing machine learning applications in industry, and internally.
Larger corporations, many of which have the need to collect and analyze hundreds of thousands or more records at a time, have little interest in consumer AI devices. They need more sophisticated capabilities such as natural language processing, advanced speech recognition, image analysis and deep learning.
"If I am building a predictive model to help me forecast market demand for a mission-critical product, I doubt there will be a service available for the Echo that will help me much with that," said one IT professional with a large manufacturing company in Minneapolis.
Many IT shops looking to dip their toes in the AI waters are choosing either to integrate machine learning technologies into existing products and services or to use it to create new ones. The machine learning category is made up of a wide assortment of development tools and hundreds of algorithms that can help IT professionals analyze data, create predictive models or identify patterns within that data.
Some companies are using machine learning business applications to create predictive models they can then integrate into an existing business process. For instance, one auto insurer is using machine learning to process claims faster and more efficiently so the company can automatically predict and estimate the cost and complexity of a claim.
Mike Gualtierivice president and principal analyst at Forrester Research
"By better calculating the cost and complexity of a claim, we can use that to automatically route it to the most appropriate claims adjuster," said a systems administrator with a Rhode Island-based car insurance company who asked for anonymity. "The [machine learning] technology makes more sense for us now because, over the past few years, we have been able to collect more data we can analyze from things like mobile devices," she said.
But it is a lack of collected data that holds many companies back from investing in AI or machine learning. Very few IT shops, even those among Fortune 1000 firms, have enough data captured to make a serious investment in machine learning technologies worthwhile.
"For machine learning to be successful, it relies on working with a rich data set," Forrester's Gualtieri said. "Users also need the necessary compute power to create these models. The good news is Moore's Law is still at work. And, more importantly, the cloud is widely available so users don’t have to invest in supercomputers and expensive infrastructure," he said.
Driven by demand among its larger corporate accounts, IBM has also invested heavily in machine learning business applications as a way to improve its predictive analytics capabilities.
"The first applications where we are seeing machine learning having an influence is in business processes," said Adam Kocoloski, fellow and CTO of Cloud Data Services at IBM. "We are focusing our efforts on not just the sophistication of a particular machine learning routine, but on the lifecycle of that particular model's training and deployment. We will be investing heavily here," he said.
Another area that could be a rich opportunity for AI in business applications is predictive analytics that can be applied to matters such as customer churn or employee retention, something large banks and insurance companies must continuously deal with.
Machine learning styles
Artificial intelligence, machine learning and deep learning are sometimes used interchangeably, but they do not mean the same thing. Artificial intelligence is typically used as a broad term to describe the ability of a computer to simulate human intelligence processes. Machine learning is a specific approach to AI that emphasizes a computer system capable of learning by finding patterns within a dataset and then adjusting its approach. Deep learning is an approach to AI that aims to emulate the human brain's approach to learning with a neural network in which data from each node (neuron) is weighed to deduce a probable result.
"People have to understand the factors that can be measured by predictive [analytics] that are causing employee dissatisfaction and identify them early on," Kocoloski said. "We are seeing machine learning systems being increasingly valuable here."
Kocoloski added that yet another area where machine learning continues to play an important role is in predictive maintenance quality (PMQ). He cites General Electric, a major manufacturer of jet engines, as an example of a large company betting that machine learning and cognitive computing will play a critical role in PMQ.
"Unplanned maintenance in many industries is very expensive," Kocoloski said. "If you fail a pre-flight check and have to bring in another plane, you have passengers delayed and flights that need to be rerouted. You would much rather catch those things during a routine or scheduled maintenance exercise," he said.
AI-powered IT tools on the horizon
While there is heightened interest among many corporate users for what AI-based business applications can do for them, only a small subset have actually bought and implemented them.
In a survey by Forrester among business and technology professionals, 58% of the respondents said they are currently in the process of researching AI technologies. However, only 12% said they are using an AI system, with 14% saying they were still in the process of training an AI system.
Another practical AI-based product available now is from IPsoft, which uses three different products working in concert to increase IT productivity. At the heart of the platform is IPcenter, an IT management platform that automates a wide variety of tasks across an organization, including help desk functions.
The second offering, Amelia, is an AI platform capable of natural language processing. The third product is Apollo, an autonomic and cognitive-enabled automation platform that uses analytics to capture a view of which tasks need to be carried out and creates an automated workflow for monitoring and tracking those process flows.
The integration of Amelia into Apollo, for instance, allows the platform to work directly with other IT management offerings, such as BMC Remedy and ServiceNow's service desk. These third-party help desk offerings connect with Amelia and Apollo much the same way they integrate with IPcenter, allowing for more advanced automations.
Amelia can support end users as well as IT staff, while the Apollo, IPcenter and Amelia combination focuses on improving IT operations. An example of this integration would be one that allowed Amelia to execute the end-to-end tasks required to close out service requests, such as unlocking Microsoft Active Directory accounts and resetting passwords, rote tasks that eat up IT time.
At the higher end of the spectrum, Google has applied machine learning applications for data center optimization over the past two years to reduce the amount of energy it takes to run its mammoth data centers by some 40%, company officials said. Google is using its DeepMind technology to achieve its goal of powering 100% of its data centers via renewable energy. This in turn will help those companies that run on Google’s cloud to improve their own energy efficiencies.
The technology giant accomplished this by using historical data collected by thousands of sensors within the data centers tracking temperatures, power consumption and cooling pump speeds. They then used this data to train a series of neural networks.
"Deep learning has the possibility of doing of lot more in the way of generalized learning on large data sets, although it is very math- and compute-intensive," Forrester's Gualtieri said. "This is what is powering Google image search and Google translation," he said.
Here's why you won't lose your job to AI
Ready for more? Read a guide to AI tech
AI may be vital in future cyber security