What Toyota, Schneider National, PayPal, and Palantir Got Right

Michael Watson Ph.D Partner
Read Time: 3 minutes apprx.

As previously published in SC Digest.

In my class on Operations Excellence, we briefly cover the rise of Toyota and all the innovative operational practices they gave us. In retrospect, we can cite the lessons from Toyota from Lean, to set-up reduction, to variability and waste reduction, and on and on.

But, at the time it wasn’t clear what Toyota’s secret was. The US Automakers saw that Toyota was up to something (by their ever-rising market share and clear leadership in quality). During plant tours of Toyota, US executives noticed clean plants and well-functioning robots. And, GM’s CEO at the time drew the conclusion that automation was the way to go. He even helped coin the phrase “the light’s out factory.” You could have a fully automated factory that didn’t need people and you could even turn out the lights, because no one needed to see anything.

It turns out that the “light’s out” factory was a bust. And, it turned out that full automation wasn’t Toyota’s secret. Toyota’s secret was autonomation– “automation with a human touch.” That is, Toyota worked hard to figure out how to best marry up what people were good at with what robots were good at. (A good source of additional information is the book Factory Physics).

This lesson is not just for the factory floor. Tom Davenport did a research report on truck dispatching at Schneider National.

The manager of the project at the time, Zahir Balaporia, said that they keep automation as a stretch goal just to see what can be achieved. However, the practical goal is more augmentation. They determine what the automation (optimization in this case) excels at– dispatching the easy loads with no exceptions and with clean data, and what people are good at– dealing with messy data, weather interruptions, and negotiating changes with the drivers and customers.

The dispatchers are more productive in their jobs and can focus on the bigger picture. This is a great way to think about optimization systems– design them knowing that they will be augmenting people.

And, there are even some voices in the Artificial Intelligence community stressing this point. This contradicts the common thinking that the goal of artificial intelligence is to simply automate. Peter Thiel points out in his book Zero to One that PayPal and later Palantir used Toyota’s idea (although he didn’t make the connection to Toyota).

At PayPal, fraud was costing the company $10M a month and preventing profitability. To tackle the problem, they first fell into the trap of trying to fully automate it. They had some success, but the fraudsters would quickly adjust tactics and get around the new algorithms. Then, in the end, what worked was to let the algorithms sort through the almost uncountable number of transactions and flag the suspicious ones and let people figure those out. This turned out to be a winning strategy and led to profitability at PayPal.

The FBI was even interested in their fraud detection system. Thiel and others realized they could use this same technique for other large scale problems. They set up the high tech company Palantir with this philosophy of combining the best of artificial intelligence with the best in human capability. They’ve used this to help take down terrorist networks (they won’t say whether they helped find Osama bin Laden, but they are always quick to point out that others have said this about them– this is either an indirect way of taking credit or stealing credit from someone else), take down a large child pornography ring, and predict where insurgents would plant IED’s in Afghanistan.

I think all of these stories reinforce how we should think about technology in operations and the supply chain. You should figure out how to use technology to do what it does best and use people to do what they do best. Having 100% automation is not ideal and having 100% human decision making isn’t ideal either. You want to find the right mid-point for your business.