Opinion: AI disruption is bleeding into white-collar jobs

Photo credit: Russian Burlaka.

A recent McKinsey study has put a different slant on the impact automation will have on jobs. It detailed a wide spectrum of jobs and identified the level of activities within each job that can be partially automated. According to the study, 30 percent of the activities in 60 percent of jobs can be automated—including even complex white-collar jobs.

The study also revealed that jobs in Japan, India, and China have the highest potential for automation, as these countries skew towards manufacturing, accommodation, and food service industries. So rather than a black or white classification of which jobs could be completely automated, the study put each job on a “continuum of automation.”

The more valued human roles are those that require non-routine cognitive work, where human interaction, problem-solving, and creativity are not readily replicable by a machine—at least not for now. These roles have fewer tasks that can be automated (think chief executives, statisticians, and psychiatrists), but I want to explore cases where even their more intelligent tasks, the non-routine cognitive ones, might be absorbed by AI so that these workers can free up their time for higher level work.

At some point, the percentage of automation even among these professions will become significant.

An everyday example

Imagine a scenario where a product manager at a telco wants to build a business case for a price change or to launch a new concept. Typically, they would weave together this business case based on various inputs: current and past business performance, the competitive environment, econometric forecasting, and customer insight. You can envision the amount of work required to collate this and the level of human expertise needed to synthesize the information. In fact, many of these tasks might be outsourced to consultants, analysts, and project managers. However, they still require human expertise and judgment.

This scenario, re-imagined, would mean a marketer or product manager could simply brief a bot (a form of AI) on their objectives and all the relevant information needed for their decision could be developed within minutes. So for those of us working in these non-routine cognitive jobs, can AI inform business decision making more generally via intelligent insight and recommendations?

The intelligent bots

We are at the stage where some of the above tasks can be achieved by AI, even if they are performed at the more elementary level. There are already some bots that have the capability to write news, analyze data and provide a simple report, and conduct preference testing for market research. There are even some in development that will be able to review web and print sources, pull out only relevant articles, summarize them, and then produce a daily report.

And there are bots like X.ai, Clara, and Evie—more efficient, accurate personal assistants at a fraction of the cost. So you can see that bots that are capable of taking on some of our “cognitive lifting” already exist.

I want to talk more about my experience with two of these bots and how they may change the way a business manager or marketer works.

Ask Sora

The first is Ask Sora, a Facebook Messenger bot capable of collecting customer insights.

This is how it works: its primary role is to help people make decisions. There are two key groups of users: askers and voters. Askers simply ask a question like, “What should I have for lunch? Chicken rice or beef salad?”. Voters then vote on what they prefer. It is a very basic premise, but it is, in essence, the cornerstone of any market research: asking people their preferences. I have been using this chatbot recently and I find that as an asker I get my results generally within an hour.

I believe this simple, intuitive bot will have commercial benefits once they reach a critical mass of users. Businesses and marketers are often looking to test new ads, concepts, and price points quickly with consumers but can be stalled by long turnaround times from traditional research firms.

Now, marketers can open up this app, upload concepts or price points, and go to a specific audience for their votes. Within an hour, a response will come back telling the marketer which concept was the winning one. And they could probably look at the results by specific segments if they wanted to (This is just a conjecture, as this seems like a basic request).

Wordsmith

A trusted advisor