Artificial intelligence (AI) has progressed from fantasy to reality at enterprises across a wide variety of industries. The total installed base of devices with AI will grow from 2.7 billion in 2019 to 4.5 billion in 2024, according to ABI Research. Global spending on AI systems will hit $97.9 billion in 2023, two and a half times the 2019 spend, according to IDC’s Worldwide Artificial Intelligence Systems Spending Guide.
“The use of AI and machine learning (ML) is occurring in a wide range of solutions and applications, from ERP and manufacturing software to content management, collaboration, and user productivity. Artificial intelligence and machine learning are top of mind for most organizations today,” David Schubmehl, research director of Cognitive/Artificial Intelligence Systems at IDC said in conjunction with the guide, noting that AI will be the disrupting influence reshaping entire industries over the next decade.
It’s clear that AI will be on everyone’s roadmap soon. As Enterprisers Project noted in our Harvard Business Review Analytic Services report, An Executive’s Guide to Real-World AI, “Hype in tech is nothing new. What’s different this time is the degree to which reasonable and knowledgeable people believe that there is, indeed, a real urgency to get going with AI now.”
[ Get our quick-scan primer on 10 key artificial intelligence terms for IT and business leaders: Cheat sheet: AI glossary. ]
That makes it even more important to get a handle on the basics. Let’s dig in:
How AI works
Typically – and at the most basic level – AI analyzes historical data to make predictions about the future.
“AI matches the data about circumstances in a business or technical process to its outcomes so that developers can make desirable outcomes more common using the people you have,” explains Whit Andrews, vice president and distinguished analyst with Gartner.
[ See also: How to explain AI in plain English.]
The AI on your phone, in your car, or in a new enterprise IT app ingests those massive volumes of data and produces an implicit model that best optimizes some predetermined objective or reward function.
“Usually that objective is represented by some calculated measure of how often the AI is correct in its outputs," says Timothy Havens, the William and Gloria Jackson Associate Professor of Computer Systems in the College of Computing at Michigan Technological University and director of the Institute of Computing and Cybersystems.
AI learns very much the way humans do – except at an exponentially faster rate.
How AI and big data work together
AI is best matched with problems for which large amounts of historical data already exist, or for which data can be simulated quickly. Today there are many types of business problems for which lots of data exists: Think customer service, cybersecurity, fraud detection, marketing, network management, predictive maintenance, and supply chain automation, for a start.
IT leaders can look for critical business junctures, particularly those problems that have a large amount of well-defined and authoritative data, and where the outcomes that they want to affect are embedded in the data, says Andrews.
In situations where you can ultimately measure the impact of the AI – whether a customer had a positive or negative experience, whether a client purchased the tool or abandoned the buying process, whether the aircraft engine was actually in need of maintenance or the prediction was off – the system can actually improve over time.
AI, when applied appropriately, can fuel much better insights from big data. For more on this, read our related article, How Big Data and AI work together.
4 common types of AI
While AI is often used interchangeably with terms like machine learning or deep learning, the latter two are subsets of the broader category of artificial intelligence. The most common types of AI that IT organizations might explore include:
ML is a branch of AI that empowers computers to self-learn from data and apply that learning without human intervention. When facing a situation in which a solution is hidden in a large data set, machine learning is a go-to.
This branch of AI (a subset of ML) tries to mimic the human mind. Deep learning uses so-called neural networks, which “learn from processing the labeled data supplied during training, and uses this answer key to learn what characteristics of the input are needed to construct the correct output,” according to one explanation provided by deep AI. “Once a sufficient number of examples have been processed, the neural network can begin to process new, unseen inputs and successfully return accurate results.”
Natural language processing (NLP)
When you take AI and focus it on human linguistics, you get NLP. It’s the branch of AI that enables computers to understand, interpret, and manipulate human language. NLP itself has a number of subsets, including natural language understanding (NLU), which refers to machine reading comprehension, and natural language generation (NLG), which can transform data into human words.
Computer vision
Computer vision helps machines identify and classify objects – and then react to what they “see.” Computer vision can learn to view and interpret the visual world in much the same way humans do – and as AI capabilities have advanced, computer vision has enabled machines to gauge things people can’t, such as temperature or air quality. Incorporating deep learning, computer vision tools get better at detecting patterns in images or other data over time.
[ Do you understand the main types of AI? Read also: 5 artificial intelligence (AI) types, defined. ]
AI strategy 101
Crafting an effective AI strategy is critical to the success of these efforts. As Justin Silver, manager of data science and AI strategist at PROS, recently pointed out, shaping and implementing a successful artificial intelligence (AI) strategy requires thoughtful analysis of business problems, care with data, and a strong organizational culture.
Those organizations getting traction with their AI initiatives invest time in identifying the right use cases and how they will measure value, cultivate data and data-related processes that support their AI initiatives, and put the right people in place to foster a culture that encourages creativity but provides structure.
It’s also important to note that AI doesn’t fit neatly into the same processes and approaches that the IT organization has used in the past. It’s a different animal. The best practices and common-sense approaches that apply to evaluating, testing, implementing, and scaling non-learning systems may not always translate. In some cases, they may backfire. In fact, we recently shared 8 counterintuitive AI strategy tips that buck common wisdom but boost AI effectiveness.
Indeed, AI is already beginning to reshape IT in a number of ways, from demanding a closer working collaboration with enterprise data science professionals to the need for developing clear and effective AI governance practices.
Where to begin with AI
The first AI steps are perhaps the most important: A promising initiative can win over skeptics while a disappointing effort can prevent further investment. So it’s important to vet early AI investments well and pilot wisely.
We have offered five questions to ask when identifying AI opportunities in your organization. Once you’ve answered those, consider these five criteria for selecting an AI pilot project.
It’s also necessary to lay the proper groundwork for AI initiatives. IT leaders need to walk through the processes of explore business opportunities, assessing data needs, examining infrastructure, determining talent or vendor needs, and preparing for inevitable risk: See Are you ready for AI? 5 places to prepare now.
How to succeed with AI
While it’s still relatively early days for many AI efforts, some emerging best practices can help IT leaders as they proceed with their initiatives. Chief among them is having a clear strategy, as noted above. In fact, 65 percent of AI high performers report having a clear data strategy, according to McKinsey & Company’s recent global AI survey.
Successful AI teams tend to be multidisciplinary, with representation from not only IT and data science but also information architects, user experience experts, and subject matter experts from the business.
These teams cast a wide net of use cases to determine how far the new AI-enabled solution might go – and help the organization identify which use cases are going to offer the quickest payback. They apply AI to solve very specific business problems. They tend to have strong executive sponsorship and support. And they focus on user adoption and experience. (For more, check out Artificial Intelligence (AI): 8 habits of successful teams.)
What do these successful AI teams avoid? Above all, they advocate against seeing AI as a hammer for every nail in the business. It’s best suited to certain types of business challenges. What’s more, applied AI is in its earliest stages of development. “AI is not a panacea to solve the entirety of the world’s problems,” Havens says. “Rather, it’s a technological toolkit for specific challenges that can be addressed through an AI-augmented approach [when it is] worth the investment to do so.”
[ How can automation free up more staff time for innovation? Get the free eBook: Managing IT with Automation. ]
