We believe in using the power of machine learning to complement our own power. The combination of skills – the remarkably fast ability of AI to accurately recognize patterns and make decisions, and the creativity of the human mind – is what will ultimately revolutionize industries. Machines cannot change the world alone, a framework that supports human and computing systems collaboration is required.

Understanding and implementing that structure, however, can be difficult for companies as they begin to adopt AI. According to a report from Boston Consulting Group and MIT Sloan Management Review, one of the top AI managerial challenge will be “figuring out how humans and computers can build off each other’s strengths.”

The same report shows that 75 percent of the executives surveyed think AI will enable their organizations to move into new businesses, and 85% believe AI will allow their companies to obtain or sustain a competitive advantage. Yet, the reality is, according to Forrester Research, even though over half of companies today are investing in AI, more than half of those have no tangible business results to show.

Even more specific, Harvard Business Review survey respondents report that the majority of ambitious “moonshot” AI projects are less likely to be successful than the more opportunistic ones which enhance fundamental business processes.

There appears to be a widening gap in terms of what machines are expected to do versus what they actually can do unaided. Forrester has gone so far as declaring the AI “honeymoon” as over, noting that machines are not the panacea for solving business problems. They require the ingenuity of humans who set the algorithms and apply the insight.

Use case

This notion is evident in our work, as we solve tough, real-world problems for our customers. For example, machine learning helped us create our Vehicle Passenger Detection System — computer vision technology that automates occupancy enforcement in HOV lanes. Traditionally, police officers manually identify the number of passengers in passing vehicles from their patrol car. Unfortunately, that practice can put officers at risk as they pursue violators across lanes, sometimes in heavy traffic, and limits enforcement to hours and locations where officers are available.

Using video analytics and geometric algorithms, machines initially learned to detect whether a seat is vacant or occupied (without using facial recognition). The machines’ capabilities to solve the problem, however, stopped at the point where uncontrolled conditions such as vehicle type, weather and lighting influenced output.

More and more data could have been added to the algorithm in the hopes that the machines would learn to overcome the conditions, but that approach would not have rendered the right results. We had to reframe the problem, rethink the approach and retrain the machine in order to be successful.

The resulting system, built through collaborative work between humans and machines, is 95 percent accurate at speeds ranging from stop and go to 100 mph. It is data-driven and enabled by machine learning augmenting a human idea.

Machines and the future

Linking what’s possible to what’s practical for our customers — that’s what we do every day at Conduent Labs. And we’re taking advantage of machine learning to transform business processes, create entirely new applications of technology and solve day-to-day business problems. Machine learning will undoubtedly play a significant role in the future of business, but I can’t yet envision any circumstances where we’ll just leave everything up to systems.

The technology is here — computing power and speed to process and analyze massive amounts of data, systems that can improve themselves over time. But investing in AI technology isn’t enough. Forrester concurs, predicting 2018 to be a year when organizations looking to get ahead with AI will invest in people — “new roles” to make the technology work.