© 2019 SourceMedia. All rights reserved.

Seeing beyond the hype to the promise of artificial intelligence

Every new technological development goes through the stage of being perceived as the silver bullet for everything: generally speaking, the less people understand the tech, the more they’ll talk about how it's going to change everything.

It happened with cloud, with mobile, with social, and most recently, with big data. Remember when the term "big data" was new and it was used as a catch-all idea for every kind of unstructured data, from streaming sensor data to photographs and documents? I haven't heard serious CIOs talk about big data that way in two or three years, because we've moved on to the next stage of the cycle.

Artificial intelligence (AI) has been following the same hype cycle as other technologies before it, but I think it's about to make that leap from an all-encompassing buzzword to something with real meaning for business. In fact, McKinsey Global Institute released a study in late 2018 which found that 47 percent of companies world-wide have embedded at least one AI-backed capability in their business processes, up a whopping 27 percent from 2017.

AI systems have potential to take on the mundane tasks that are too time-consuming and therefore too expensive for human labor, and/or too prone to human error. Businesses are already developing robotic process automation (RPA) solutions capable of doing these tasks.

It won't be long before it's easier and more affordable to hire or buy a bot to scan documents or match purchase orders to invoices. It will also be more humane than forcing low-level staffers or outsourced workers to bear the responsibility if an error in a repetitive, low-thought task has significant consequences.

The use of AI will soon expand to support enterprise decision-making. Some companies have already reached this milestone and are reaping the rewards, such as the delivery and carrier companies that use AI to optimize delivery routes in real time – a fairly mature application.

AI jobs.png

UPS, for example, saves an estimated $50 million a year with an AI-powered tool that creates the most efficient routes for its fleet. However, this type of AI deployment, like any other, is only successful if it's driven by data that's relevant, trustworthy, and up-to-date – a proofpoint for the critical importance of a data governance program.

As companies experiment with AI, they'll need to figure out how to scale up the successful experiments for the entire company and make sure they work together. The hardest part of introducing AI will not be deploying a hundred AI-driven applications, but getting all of those applications to share data. That requires a common data structure with shared definitions of key terms.

As companies consider how to incorporate AI into their operations, a few things to consider:

1. Start all of your experiments with AI with a common platform and data model. Most of your experiments will fail, but the ones that succeed will be easier to scale up and combine without having to rearchitect them.

2. Keep your expectations realistic and reasonable. When you tell the CEO you're working on an AI project, the CEO shouldn't imagine something out of a science fiction novel. Even one of the best-publicized AI projects to date, which was more accurate than doctors at identifying stomach cancer from photos, was less about AI's skill at spotting images of cancer than about bias from human beings who assessed the photos differently depending on the order they saw them.

3. Don't get caught up in the excitement of working with AI and machine learning for their own sake. IT and business need to work together to keep the focus on the business problem they're trying to solve.

4. Believe in AI's potential. Your data scientists may be used to spending hours extracting data and making it useful before they can start analyzing it to drive decisions, but that doesn't mean there isn't a better way.

5. Hire people who understand data models and can put together data strategy and a proper data structure.

Don't try to do too much at once, but don't wait for AI to reach maturity, either.

If you're experimenting with multiple small AI projects across multiple functions, such as security, finance, and customer service, your IT organization will be ready to react to the next big breakthrough, whatever it is, when it happens. Because if AI follows the same path as other aggressively-hyped technologies – and it will – that breakthrough is just over the horizon.

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