3 new jobs A.I. is creating: Trainers, explainers, and sustainers | Paul Daugherty

One of our fundamental premises with 'Human + Machine' is really the “plus” part of human plus machine. There’s been a lot of this dialogue about polarizing extremes, that the machines can do certain things and humans can do certain things, and as a result we end up with this battle, kind of pitting what the machines will do versus the humans. We think that creates the wrong dynamics.

So with 'Human + Machine' we’re trying to reframe the dialogue to: what’s the real interesting space, and really the big space, where humans and machines collaborate—we call it collaborative intelligence—and come together and help provide people with better tools powered by A.I. to do what they do more effectively? And if you think about it that way, we really believe that with A.I. we’re not moving into a more machine-oriented age, we’re actually moving into an age that’s a more human age, where we can accentuate what makes us human, empowered by more powerful tools that are more humanlike in their ability, and that creates these new types of jobs.

So we call that the 'missing middle' because there hasn’t been a lot of discussion about these jobs in the middle where people and machines collaborate. And we’ve come up with two sets of jobs. On one side, you have the jobs where people are needed to help machines, and that’s not a category that too many people focus on. We think it’s an important one and I’ll come back to that in a minute.

On the other side, we have a set of jobs where machines help people, machines give people new superpowers. And those are the two broad categories of jobs we see in the 'missing middle'. So in that set of jobs where people are needed to help machines, there are a few interesting, novel, new categories of jobs we found that people don’t often think about and we call those trainers, explainers, and sustainers, and they’re very important things for all organizations to think about as you think about how to deploy artificial intelligence in your organization.

So think about a trainer. What we mean by a trainer is it’s a new type of job where a person is needed to train A.I. or train the machines that we’re using in businesses. We’re not talking about simple things like tagging data for supervised learning—that’s included, but that’s just the start of it. What we’re really talking about here is more sophisticated forms of training that are needed so that our artificial intelligence and our systems behave properly.

For example, for companies we’re working with that are developing chatbots and virtual agents, if you’re a bank you might want a very different type of personality than a media company or a gaming company or a casino, and embodying the personality, the behavior, the culture, the characteristics, the nature of the response in your A.I. is a really important consideration for companies. Because we talk about the idea that with A.I., you know, A.I. becomes the brand of your company because it’s the face of the company and how your company is perceived by your customers.

So this idea of a trainer that brings in skills to develop that kind of behavioral response for your A.I. is a really important skill. And we’re hiring people to do these jobs today, people with backgrounds in things like sociology, psychology and other areas. Not a technical skill but a new type of role that’s very important to get A.I. right as you apply it to your organization.

Another type of job where we see people needed to help machines are explainers and sustainers, and I’ll talk about these two a little bit together. Explainers are new roles where we need people in roles where they can explain the implications of artificial intelligence. One thing you hear a lot about is 'black box' A.I.: “AI can’t be explained. We don’t know how it’s operating.” Yes, that’s true for certain forms of A.I., but we need explainers to help us decide: when is it appropriate to use black box A.I. and those types of algorithms, and when do you need to be able to explain how you arrived at a decision?

In every organization you’re going to have cases where you need to explain and cases where it’s okay to use more of a black box algorithm. We need explainers to help organizations make those decisions. And then also explainers to step in once a decision has been made to explain the business consequences of it after the fact. So an explainer becomes a very important capability.

And then sustainer is also important. Sustainers are looking at: what are the ongoing implications of using A.I.? Is the A.I. performing as expected? What are some of the rules and policies around how to use A.I. in your company? And with some of the things that are in the headlines today that we see in recent weeks—things like the tragic circumstance with Uber, with the person being killed tragically in Arizona.

Or Facebook in the headlines over what turns out to be legal use of their information under their policy, but a policy that’s being questioned. These are areas where companies need explainers and sustainers to think about not the implementation of the technology but the appropriate use of the technology, the transparency of it, the ability to explain it, and that’s what we’re talking about with the explainer and sustainer roles.

And this isn’t something just for tech companies. Every company is developing digital platforms. They’re using data in novel ways. They’re developing increasingly intrinsic consumer services, customer services and experiences using artificial intelligence and other technologies. So every organization needs to embrace and understand these new roles and these new issues involved in applying A.I. to their organizations.

Now the other side of the 'missing middle' is where machines help people and give people superpowers as we talk about. And there are three categories of jobs we identified here as well based on our research. There are jobs where machines or A.I. amplifies people, so we call it 'amplify'. There are jobs where machines aid in interaction, so the 'interact' category.

And then there’s 'embody', which is actual physical interactions: robotics and cobots and such. And those are the jobs where machines and A.I. really empower people and allow them to do more. To give you an example that we see here that’s interesting is: think about the 'amplify' category. One great example is Autodesk which is design software, AutoCAD, that designers use to do software. It used to be it was very much like engineering drawings, where you had an engineer or designer specifying exactly what they needed.

Well, with DreamCatcher, a generative design software using A.I. that Autodesk delivers with AutoCAD, designers can now use A.I. to multiply the power of their creativity. A.I. can amplify their own creativity, come up with thousands of designs based on constraints and parameters that the designer sets, and then the designer’s own creative skills come into play to curate and select and modify those designs to produce a truly unique, human and creative output.

And that’s an example of software and A.I. amplifying the capabilities of an individual, it's a great example of this giving a designer, in this case, the power to do more and do more creatively, leveraging their human skills. You know, a great example of interact and interaction skills where machines can help people is what we see in customer service. We’re seeing a widespread use of virtual agents and chatbots and A.I. technologies like that to help automate and better serve customer needs in many, many industries.

But a really interesting use case we’re seeing in many industries is how A.I. and these chatbots can help people serve customers more effectively. How we can use virtual agents to automate the mundane tasks so a customer service worker doesn’t need to spend time typing on a keyboard but can spend time empathizing and communicating with a customer.

One example that we’ve seen is in a banking environment where there’s a lot of compliance rules. And an agent may always be thinking about what are the compliance restrictions, regulations, and looking at their computer and deciding how to do all the compliance as they’re interacting with a customer.

Well, that’s a case where companies are applying virtual agents that are kind of the wingman for the customer service agent, listening to the conversation, watching the transactions, understanding the compliance implications and helping the agent focus their time on the consumer and offloading the compliance-driven obligations.

It’s a great example again of the machine partnering with the human to allow the human to do what they do best—empathizing and interacting with the customer—while the A.I. takes on compliance and other transactional activities.

And finally, 'embody' is where you see people working together with physical robotics, so it’s embodying the physical capability of people along with robotics. One example that we talk about in the book that I think is a powerful illustration of this is Mercedes with its new S-Class factory.

The S-Class vehicles—Mercedes used to be heavily automated, over 80 percent automated using large-scale, industrialized robots for the factory. What they found was that was too inflexible. It didn’t allow them to be adaptive and flexible and personalized enough to meet the demands of their customers. So Mercedes kind of reverse-engineered the factory, brought more people into the process, equipped the people with cobots that people could dynamically adapt, and added more people to the equation with cobots so that they got more personalization, better productivity in the factory than they did in the heavily automated sense.

So that’s a great example of using advanced technology, artificial intelligence, robotics combined with people to produce an outcome that was much better for the business, in this case, the manufacturing process. One of the fusion skills that’s really interesting is judgment integration. Judgment integration is talking about how do we combine the judgment of a person with the judgment of a machine, and combine it together to make the best decision between the two of them?

An illustration of why this is important comes from a recent Harvard medical study that they did. They looked at the accuracy of breast cancer/breast tumor detection. It turns out that A.I. right now can detect breast tumor conditions with about 92 percent accuracy. Physicians can do that at about 96 percent accuracy. If you put the two together and give a physician access to the results from the algorithm it’s 99.5 percent accurate.

So it’s the combination of the human plus the machine gives you the best outcome, and we find this in profession across profession.