The Future of Human-AI Collaboration: From Survival to Success

Alex Dichter is Senior Partner Emeritus at McKinsey & Company.
Anna Jaffe is the Co-founder of Mobi.AI.

We as a species have evolved to survive. For good reason, humans have been focused on how to live in the safest possible way throughout time. This means that, as a general rule, we have had to put survival first, not excellence. But we are entering a time where technology, and human-collaborative AI in particular, can give us the ability to discover and choose the more difficult paths we would normally avoid, showing us the way to get to the outcomes we truly want.

The world that we live in today is so different in scale from what we evolved to be able to mentally manage. It can take weeks for fruit fly generations to evolve new biological characteristics in response to change, but in humans it takes far, far longer. It’s striking that we suddenly have the entire world available to us through our phones, yet our brains are still hardwired to be afraid of the animal predators in the dark. We still think about coordinating logistics at a village scale, rather than a global scale. Our thought processes simply aren’t yet adapted to the world we’ve built and live in every day.

However, we now have a set of tools that can aid us in this struggle. Human-collaborative AI can work together with humans to solve real problems in real time, at scales we could never do on our own, while playing to our strengths so that we can steer.

Alex Dichter speaking to the Mobi team.

We recognize these are some broad statements, so let’s give some specific examples from one of the industries we have expertise in: travel. How could airlines show the way for the future of human-collaborative AI? And in light of the recent CrowdStrike event, how could human-collaborative AI have helped?

Let’s start simply. Imagine a gate agent—let's call her Michelle—deciding whether or not to hold a flight for a connecting passenger from an inbound delayed flight. From her point of view, all she sees is the plane down the ramp fully boarded and ready to go on time. All she hears is a call to deliver an on-time departure. All she wants is to survive that step in her day. She can’t see the passenger running to her (or know how fast they’re likely to run). She can’t easily see the downstream impacts of holding the flight, or the impact of her decision.

Even for this simple example involving, at the surface, two people, humans can’t solve a problem with this many variables, but human-collaborative AI can. By using a full suite of different AI tools—analytical AI, neural nets, search, expert systems, operations research tooling, behavior economics models, digital twins—and others. For example, the large language model (LLM) can understand, based on all of the walkie talkie conversations at the original departing gate, that the flight is likely to be late before that information shows up in the system today. The routing layer can determine, based on their seat number, the actual path between gates, and the neural net can predict how fast they are likely to walk and whether they might stop for coffee even though everyone is waiting. It can know how likely the specific jet bridge operator is to be sitting in the seat as the plane pulls up, and how quickly they are able to connect. The operations research layer can trace the impact of holding the plane five minutes later on any onboard crew, the passengers, and their connections, and that plane’s next flight. The network monitoring layer can check to see whether there is air traffic and thus if holding the plane by five minutes would actually mean that it would have to be held for an hour due to weather, or lose its spot in the queue. This network monitoring layer could then determine if the delay would mean that the crew would have to clock out, meaning the flight wouldn’t be able to depart at all.

Now what’s the job for the human in all of this? We are now freed up to do the things we’re great at. We’re never going to run out of work. The person can make the final decision to hold the plane, send the plane, and/or put the person on a different flight, all impacts clearly shown to them. In this case Michelle gets to be the expert. She’s also free now to focus on the passengers in front of her rather than trying to solve a network optimization problem her brain never evolved to solve.

Stepping back, the power of human-collaborative AI is not only to help us solve problems, but to allow us to excel in a world that has expanded beyond our mental scope. Whether aimed at travel logistics, deciding how to manage business resources, or even how to best provide aid during disasters, human-collaborative AI can give humans the super power of expanding the scale of our ability to see variables and best paths forward in decision-making. But we need to feel comfortable working in collaboration with AI, even when it recommends solutions we would have never considered. It’s important to challenge the things we know to be true, because as we have seen the world is going ahead and changing all of the time. AI can help us not only keep up with these changes, but excel during them.

Historically, technology was really only five percent better than any expert person. One of the reasons for this small level of improvement so far is that when engineering these technologies, people were leaving out 80% of the problem itself. As we saw in the airline connection example, there are so many details and facets that are actually a part of the equation beyond the obvious. If you can’t look deeper than the surface layer, then you aren’t truly challenging your assumptions about how to solve it. The other half of the issue that would push our technological innovations forward is that the technology actually has to collaborate with the human expert. If you can do those two things—deeply model the problem and improve collaboration between AI and the expert—then you can solve business-critical issues in a way that could never be done before. We think that’s the wave that’s coming: the next phase of the era of AI and it’s human-collaborative.

If we are intentional and thoughtful about how we shape the tool we’re creating, then we are being intentional about the path we want into the future. If we wanted AI to replace us, the tools we would build would be very different compared to if we want AI to play to our strengths and collaborate with us. A future we choose to build with AI will be totally different than a future we allow AI to build for us.

The Authors

Alex Dichter spent 25 years leading McKinsey's Greater Boston Office, and also led the Firm's global Airline, Travel and Aviation practice. Alex’s leadership of McKinsey’s Travel, Logistics & Infrastructure Practice entailed supporting clients on many fronts, from strategy development to regulatory reform efforts. As is fitting for a pilot, Alex has assisted airlines and aerospace companies around the world, most recently in bipolar efforts in Canada and Chile. Prior to joining McKinsey, he spent 8 years at Continental AIrlines as a pilot, instructor, and check-airman.

Anna Jaffe is the CEO of Mobi.AI, where her technical expertise as an MIT graduate and industry leader has shaped her vision to use technology to solve large-scale, intractable problems. Her academic research focused on biomimetic design and using natural systems to help people solve the most pressing issues of our time. She is particularly interested in the intersection between technology and humanity—and how technology can increase our empathy, sensitivity, and decision-making capabilities, especially in high-risk situations. Anna’s passion for exploring how AI can shape the future for the better is defining how Mobi.AI impacts the world.