The great unknown: where's it all leading.

getty

I once heard that, hypothetically, if the telephone company didn’t modernize its national phone network in the 1950s – which was all connected by human operators manually plugging switches to complete phone calls – the phone company would need every single adult in the country to cover its needs. That’s a pretty good case for automation.

That analogy came to mind as I was speaking with Jon McNeill, CEO of DVx Ventures, former president of Tesla, and former chief operating officer of Lyft. As the phone company automated its network, what followed was a great deal of “hand-wringing with fears that 800,000 people would be out of work,” he recounted.

That was the first-order effect, he explained. “What people couldn’t see at the time was the backside of that. Entrepreneurs looked at that and created a whole new industry called call centers. We ended up having many more jobs, much more GDP creation and value creation."

In his new book, The Algorithm: The Hypergrowth Formula That Transformed Tesla, Lululemon, General Motors, and SpaceX, McNeil discussed the impact of technology on businesses and markets, how technology drives opportunities that are difficult to see right away. At the same time, he outlined why automation can’t be applied willy-nilly either.

McNeill pushes back vigorously against the notion of a job wipeout due to AI. In the long run, technology developments such as AI will deliver new generations of innovation – and create new roles for humans. “I hear the argument, ‘this time, it’s different,’ that ‘the invention’s smarter than us,’” he said. “Inventions are always better than us. I just don’t see evidence that this is the first time in human history this happened. We can’t see the other side, we rarely see the other side, and the opportunity it creates.”

It’s common to clearly see “first-order job destruction in technology,” McNeill continued. “But humans are really bad at seeing second-order effects, and third-order effects. All the doomsayers can see is the doomsday scenario."

For example, he said, there used to be skyscrapers full of people doing manual calculations. The spreadsheet came along and eliminated all those jobs. But we didn’t have mass unemployment on Wall Street. It enabled all kinds of different trading strategies and second-order effects that actually led people to have much more interesting jobs."

To prepare an organization to optimize AI, it’s important to think long term here as well – maybe not as far as second-order effects, but to the effects on business processes. “First, decide what outcome you want from this business process,” McNeill advised. “Asking for AI first is a lot like saying, ‘I want a hammer.’”

The bottom line, he said, is “if people don’t understand the nature of the problem they’re trying to solve, what their goal is, they’ll start applying a hammer to every problem. They’re only going to leave holes in the wall.”

The best approach is to describe "‘here’s what I want you to build. Now let’s talk about how you go about that. Is ML the right approach? Is AI the right approach? Is deterministic code the right approach?’"

Think about what teams are being asked to build, he continued. “And start with the goal. And then let teams tell you what is best to reach that goal.” For example, an executive should say, “’here’s what I want you to build, and I want you to take 50% of our cost base out,’” he illustrated. “That’s a different order that gets placed. Then people start thinking creatively about what new sets of tools they need to do that. That can pull the AI forward.”

In his book, McNeill has several guiding principles, staring with “question every requirement,” with an emphasis on simplicity, and ending up with this surprising piece of advice: “automate last.” He related how Tesla, under his watch, was experiencing a backlog of orders due to sluggishness in its automated manufacturing process. The solution, he found, was to set up a long tent outside Tesla’s California assembly plant and hand-assemble each component in an all-human assembly line. It was there the team was able to identify how to smartly apply automation to each phase.

The key is to learn, simplify, and optimize the system first. “Automation, whether software or robots, had to come last,” McNeill wrote. He likened the process to building a house, with the contractor building it as the architect is still working on it. All too often, the software is applied before the business process is mature.

“People really need to understand, ‘here’s the problem I’m trying to solve, here’s why AI is appropriate for this, and hers the steps to deploy AI,’" he said. This may include cleaning up data, training the model on the data, and doing post-training to ensure the model delivers accurate, up-to-date results.