The Fleet: How AI and Autonomy Are Redefining Fleet Operations in 2026

Participants: 

Philipp Kampshoff, Co-founder of McKinsey's Center for Future Mobility

Dor Shay, CTO of Element Mobility and Autofleet. 

Moderator: Chris Brandt 

The Four Major Forces Reshaping Mobility

Philipp: The biggest forces we see right now are what we call the ACES: autonomous driving, connectivity, electrification, and shared or smart mobility.

At McKinsey, we track inorganic investment into these four technologies. Over the last 10 years or so, we have seen more than $1 trillion invested against these trends.

Innovation is a function of time — how much time is put against a problem — but it is also a function of resources. What we have seen is that 92% of that $1 trillion came from outside the automotive world. That has significantly increased the rate of innovation in the industry. It is one reason we have moved relatively quickly toward electrification, autonomous driving, and related technologies.

Now AI comes on top of that. You not only have the capital and resources going against these trends; you also have AI agents that can help with innovation in the future. It is almost like having 10 times the number of employees innovating against these technologies. We will probably see an even faster pace of innovation as a result.

Chris: Dor, what are your thoughts?

We see a shift from resource accumulation to optimization

Dor: Similar to what Philipp said, we see a shift from resource accumulation to optimization.

Historically, the only way to grow a fleet or serve more demand was to add more vehicles. Now the mindset is shifting toward optimizing the assets you already have.

All the technologies Philipp mentioned are ultimately about using your fleet more effectively. If you simply add vehicles, you add operational cost and more chaos to the system. But AI, automation, and connectivity allow you to improve utilization of the fleet you already operate.

Some people say you need all of these technologies together for them to work, but that is not necessarily the case.

Think about the Waymo use case. In an autonomous fleet, you do not have a driver in the vehicle. That means you no longer have the main sensor that used to detect problems, and you no longer have the person in the vehicle who used to solve those problems.

If an autonomous vehicle has a flat tire, the operation needs automation that can detect the issue, reroute passengers already in the vehicle or scheduled to be in the vehicle, and send a service unit to the vehicle location. In an autonomous fleet, this entire process needs to happen in less than a second, so downtime stays low.

The interesting part is that every step I just described can also apply to a conventional fleet. Nothing about that workflow depends on the vehicle being autonomous. As long as the vehicle is connected and the operational side is automated, you can automate almost everything outside of driving. That is available to fleets today.

How Fleet Managers Should Approach Rapid Change

Chris: We are talking about autonomy, which has come in fits and starts. Electrification has momentum, but it is not evenly distributed around the world. It depends on the market, the products, and the infrastructure. Connectivity is changing. AI is coming in like a bull in a china shop and changing everything quickly.

How should fleet managers look at all of this when everything is happening at once?

Philipp: You are absolutely right. The future is already here; it is just not evenly distributed. We see that with electrification and autonomous driving.

In China, more than 50% of new vehicle sales are now electrified. In autonomous driving, if you live in certain U.S. cities, you see it everywhere. But in large parts of the rest of the world, you do not see it yet.

For fleet managers, I would say: do not fall into the trap of jumping on every technology too quickly.

These are still new technologies, and they are not fully mature. New technologies usually sit on a cost curve. In the early years, you are still on the steep part of that curve, where costs can come down significantly each year. Only once the technology matures do you see the curve begin to flatten.

We are still seeing that with electrification. The cost of batteries and electric powertrains continues to come down significantly year over year. There are also rapid advances in how people use powertrain and battery technologies, which has major implications for total cost of ownership.

The same is true for autonomous driving. Yes, we are already seeing robotaxi applications, which is phenomenal, and the ability to remove the driver is meaningful. But if you look at the actual business case and the cost of running operations, most robotaxi operations are still highly unprofitable today because they have not yet been optimized for cost.

So far, most robotaxi players have focused on safety as mission number one, two, and three. First, they needed to make the technology safe. Only after that do they start asking: How do we reduce vehicle cost? Can we trim down the number of sensors? How do we maintain and operate these vehicles more efficiently?

For fleet managers, the key is to look at these technologies closely, understand whether the business case is already there, understand how fast it is improving, and identify what internal changes are required to adopt the technology.

Sometimes the better answer is to be a fast follower rather than the first mover.

That said, some technologies will require changes to day-to-day operations and processes. Those changes can be hard, and you cannot make them overnight. In those cases, it may be worthwhile to run a pilot even if the pilot itself is not profitable yet. The point is to understand how your operations will need to change. Then, once the technology becomes cost-effective and the business case turns positive, you know how to scale it.

Chris: Dor, based on what Philipp said, fleet managers have so many factors to account for. There is the complexity of EV infrastructure, but there are also softer factors, like using certain vehicles as incentives to keep workers happy. There is a lot of technology coming at organizations, and fleet managers are trying to evaluate which direction to go while also standardizing as much as possible.

Do fleet managers need to move quickly now? Are they not moving quickly enough? Or should they move more slowly?

the value we provide to fleets: technology that helps them stay ahead of the curve.

Dor: Daniel Kahneman said that people experience failure about twice as intensely as they experience success or gain. It is very easy to reach a point where you stop experimenting because you are counting your failures instead of counting the pilots that succeeded.

I understand Philipp’s point that many of these technologies are new. But if you want to stay open to experimentation, and if you count only the failures rather than the pilots that worked, it will be very hard to stay ahead of the competition.

This is kind of the value we provide to fleets: technology that helps them stay ahead of the curve.

The cost of experimentation includes the actual cost of the software or technology, but it also includes resources. To run an experiment, you need organizational buy-in and support. That is not easy to get. But the potential gain can be much bigger, especially if you are the first to capture that advantage.

It is easy to forget that innovation moves quickly, but competitors follow quickly as well. Innovation is not something you do once and then rest for the next year. You have to keep innovating and keep trying to stay on top of it.

Without the right mindset and without allowing yourself to fail when needed, it is very hard to remain ahead.

Data, Automation, and the Day-to-Day Fleet Operation

Chris: One of the big trends we hear everyone talk about is data. There is data in the vehicles, telemetry data, data in back-office systems, and data flowing through operations.

How do you see data and automation impacting day-to-day operations in a real-world context?

Dor: The first time I saw the impact of using data in the fleet world, it really amazed me.

I visited the operation of one of our clients and saw a few people drawing arrows on a map of the city in Microsoft Paint. They were trying to figure out where vehicles should move, from where, and when.

They were having a very intense day. Once they finished drawing the arrows, they had to call different vendors, make sure no one was committing fraud with the vehicles, and make sure everything was working correctly.

Six months later, we had completely automated that entire workflow.

The same fleet manager still sits in the same chair. He still has the same title. He comes to the office at the same hour. But his day-to-day has completely changed.

The things he used to do manually are now automated. He now has much more time to think about what new business models he can add to the fleet, how to onboard better vendors, and how to secure better rates — the types of things he never thought he would have enough time to do.

Today, that is his day-to-day. The repetitive tasks are automated, and he can focus on higher-value work.

Why Data Is a Differentiator — and Why It Is Hard to Use

Chris: Philipp, I love the idea of using data and automation to shift people into higher-value work. But the environment is also more complex. There are more nuances across vehicles, more regulations, and more things fleet managers have to worry about today than they did five years ago.

What are the biggest barriers keeping fleets from fully utilizing all the data they have? Is it a systems and practices problem, a lack of appropriate data, too much data, or all of the above?

Philipp: Probably all of the above.

First, I agree that data is already a differentiator, and it will become even more important.

When you think about AI and agentic AI, many of the large language models themselves will likely commoditize. Everyone will have access to them. So how do you differentiate yourself from competitors if both of you have access to the same tools?

That is where data comes in. Do you have your own instance that is fed by your own superior data and produces better answers for you? Or do you partner with someone like Element Fleet to do that for you?

In general, I believe data becomes the real differentiator.

And yes, complexity is increasing. Think about transitioning an internal combustion engine fleet to an EV fleet, such as parcel delivery vans.

The first question is: What kind of charging infrastructure do you need? The answer depends on many things.

What battery size will you put into the vehicles? Do you use larger batteries so you only need to charge every other day? Do you use smaller batteries to reduce vehicle cost, but then charge every night and require twice the charging infrastructure? Or do you split the difference and go hybrid?

It does not end there. It also depends on routing and dispatching. Where are the vehicles going? When do they return during the day? When can they charge? It depends on the driver and how the driver uses the vehicle. It depends on telematics and the current usage of the battery. It depends on the outside temperature. What is the weather tomorrow? Do you need to run air conditioning or heating?

All of that data exists. But you have to bring it into one place, such as a data lake, and then run optimization on top of it.

Yes, it is much more complex than it used to be for fleet managers. But it is possible. With LLMs now entering the equation, it has become easier than it was two, three, or four years ago.

The first major challenge is that fleets usually do not have all relevant data in one place. You need to aggregate the data, clean it, and make it usable.

The second challenge is that many companies make the mistake of saying, “I will hire two or three data scientists, and they will solve all my problems.” That is not feasible. It does not change the DNA of how the organization operates with new information, agentic AI, and related technologies.

This is something you have to train much of the organization on. You cannot outsource it to two or three data scientists. You can build the capability internally or work with an external partner. Both can work. But it cannot be solved by hiring just a few people.

The third challenge is the business case. I see many companies playing with these technologies almost like toys. They run a pilot, use the technology, but never scale it. They also do not take the next step of flowing the impact through to the P&L.

In the worst case, you add a tool, add new employees, add additional cost, and do not get the benefits.

So the big challenges are: get the data into one place, build the capability internally or through a partner, and drive real operational impact. Otherwise, what is the point?

Bridging the Gap Between Data Collection and Operational Action

Chris: A lot of AI projects today feel like small party tricks, and the real problem is scaling them.

Dor, you talk to a lot of fleet managers about these topics. Collecting, enriching, and using data is probably outside the core competency of many fleet managers. Without turning every fleet operation into an IT shop, how do you bridge that gap?

By integrating  two platforms and making sure the data streamed correctly, we saw a 70% reduction in fleet downtime.

Dor: A lot of fleet managers spend time and money gathering data, but transforming that data into action is the art of the process. We see clients who have similar types of data coming from multiple platforms. The data exists, but because the platforms are not integrated, they cannot unlock its full potential.

We worked with one fleet in New York that had two platforms running in production. They were paying the cloud cost and had already paid for development and service costs. Both platforms were gathering data, but there was a swivel chair between them: a human reading data from one platform and manually entering it into the other.

By integrating the two platforms and making sure the data streamed correctly, we saw a 70% reduction in fleet downtime.

That is huge. It means the fleet could utilize vehicles it already owned without spending money on additional assets.

And the cost of that integration was a one-time cost. It was not even about AI. It was about getting the basics right: making sure the right data exists and flows through the organization in the right way.

Chris: It sounds like there is a lot of low-hanging fruit in these organizations. Small changes can have a big impact, and there is still a lot of ROI to capture.

Philipp: I agree. We are just getting started.

I spoke with one autonomous taxi player recently, and they said the savings are everywhere — like fruit on a tree. They just had not gotten after them yet.

I think that is true for many AI applications as well.

Chris: A lot of organizations may not be used to bringing in outside help, specifically around technology. But if the ROI can be demonstrated, and if the opportunity is relatively short-term and practical, it seems like a strong to-do list item for fleet managers.

Dor: One advantage autonomous vehicle fleets have is that most of them are digitally native. These companies were founded recently. Agile ways of working are part of their DNA. They take small steps, measure the results, and react accordingly.

It is not one giant move. It is many small steps.

We see simulation as a key component for managing and planning fleets, because it allows you to test and measure plans without spending money or putting vehicles on the road.

We see that ourselves when we partner with new fleets. Usually, the first part of the engagement is: let’s not change your real-time operation yet. Let’s simulate it.

The Autofleet Simulator can replicate a real-time fleet operation in simulation and produce results within hours.

That allows operators to test questions like: What would happen if I used EVs, hybrids, or plug-in hybrids? What would happen if I opened a new operational center downtown versus in the suburbs? What happens if I change vehicle types, driver shifts, or the mix of autonomous and non-autonomous vehicles?

Autonomous vehicles today often operate in restricted zones, so these questions are difficult to answer without a simulation tool. With the right simulation capability, those questions become much easier to analyze.

Technology Trends Fleet Managers Should Watch

Chris: Philipp, you are talking to a lot of people across the industry. There is a lot of technology flooding into the space. That creates opportunity, but it is also complicated and confusing, especially for people getting up to speed.

If you had to identify the top technology shifts fleet managers should watch, what would they be?

Philipp: I would put them into two buckets.

We have talked about electrification and autonomous driving. I would argue that electrification, if I am being a little facetious, is “just another powertrain.” It helps with sustainability and other important objectives, but it does not necessarily change day-to-day operations as dramatically — apart from charging.

Autonomous driving, however, is a true game changer for the industry. It will drive massive changes in operations. As Dor said, you need to simulate many things once you move toward autonomous.

Here is a simple example. Imagine parcel delivery in Manhattan. Today, you have parcel delivery vans with drivers. Autonomous driving comes with the promise of removing the driver.

Now imagine an autonomous vehicle driving around Manhattan. Who still handles the last meter — taking the parcel from the vehicle into the building, managing the lift gate, or unloading the truck?

Do you still need a person in the vehicle? If so, you may not save money by going autonomous.

But maybe there is a different way to think about the operation. Perhaps the vehicles deliver parcels autonomously, and instead of having one person in each vehicle, you have people on the street with scooters. When a vehicle arrives and is ready for delivery, that person goes to the vehicle, unloads the parcel, and then moves to the next vehicle.

That is a complete change in how you run operations, and only then do you truly unlock the savings.

We modeled whether it is cheaper to own a car or use ride-hailing 100% of the time in a city like Washington, D.C. The answer was roughly 2,500 miles. If you drive more than 2,500 miles per year, you should own a car. If not, you should use ride-hailing.

But if you remove the driver and have autonomous robotaxis, that break-even point changes to about 7,500 miles. Only about 50% of people in D.C. drive more than that. So, from a purely economic perspective, for about half of the people in D.C., it would not make sense to own a car anymore. It would be cheaper to use robotaxis and rent cars for longer trips.

I am not saying 50% of people in D.C. will switch entirely. But today, the average U.S. family has two cars. That might move down to 1.8 or 1.6. I do believe we will see much more fleet ownership going forward. That is why autonomous is such a game changer.

Chris: I am surprised the break-even point is that high. Owning a car is expensive, and ride-sharing vehicles have much higher utilization. What keeps that threshold lower today?

Philipp: The biggest cost in ride-hailing today is still the driver. Depending on the city and the service, the driver may account for around 60% of the cost.

For some day-to-day use cases, that is still too expensive. A daily commute that takes 30 or 45 minutes is too expensive to do by ride-hailing every day.

Autonomous vehicles promise to remove the driver cost. Yes, some of that cost is replaced by additional operations and sensor costs. But ultimately, the idea is to bring the cost down significantly, making robotaxis much cheaper than human-driven ride-hailing.

What Fleets Should Pilot Now

Chris: Dor, companies are experimenting with many technologies, and there is an advantage to getting in early. What would you say is important to pilot now?

Dor: We see changes everywhere, and people should be excited about that.

I recently read about a suggested metric for measuring the health of society. Instead of only looking at GDP growth or the number of patents, the idea was to look at the number of companies founded and then closed within the first 24 months.

That reflects creative destruction: trying something, being willing to shut it down when needed, and clearing room for new ideas and concepts. That is an important part of innovation.

Some ideas will work for some companies and not for others. I do not think there is one universal rule of thumb or one specific list of technologies every fleet should explore right now.

For example, one of our partners runs a delivery operation where more than 83% of deliveries are below five kilograms and within 15 kilometers. In that case, drones may make sense. But drones might work for one company and not another.

Fleet technology might work for one company and not another. Battery technologies may create an impact in some use cases and zero impact in others. It depends on the use case you are trying to solve.

Fleet managers should stay up to date with the latest technologies and be ready to test them. Sometimes it will be too early. But at least you avoid losing the advantage of learning before everyone else.

Where Fleet Managers Should Be Cautious

Chris: There are obviously good things to start investing in now, but there is also a lot of hype. What would you say fleet managers should pump the brakes on, even if everyone is talking about it?

Dor: We have seen that with EVs for some fleets. Electrification can create a lot of impact, but for some use cases, it may not be there yet.

There are two categories of technologies fleet managers can experiment with.

The first is vehicle technology. That means replacing an asset you already own. That asset is expensive, and if there is a problem with it, you may need to wait for the next generation. These changes are harder.

The second is fleet technology. This is usually software or aftermarket devices. It is much easier and faster to experiment with.

Vehicle technology may increase the number of miles you can drive per charge. But fleet technology can create a shift that is just as important. It can improve the amount of value you get per mile through better dispatching, better routing, better vehicle allocation, and real-time reactions to real-world events.

That type of automation can change how the fleet operates and improve efficiency.

Chris: Philipp, what are you seeing that may not be ready for prime time?

Philipp: First, if I flip the question and ask where I would deploy technology, I would say: follow the money.

Go through your P&L and ask which elements can be improved by deploying these technologies.

Start with labor cost and drivers. That immediately leads you to autonomous driving, robotics, and related technologies.

Then look at vehicle maintenance. How can you use technology for predictive or preventive maintenance? Or how can you ensure the right spare parts are in the right location at the right time?

Then look at depreciation. How can artificial intelligence help maximize residual value? When you remarket a vehicle a few years later, which features should you include now to sustain residual value?

Then look at the back office. Agentic AI is being deployed widely. I have seen finance organizations lean into agentic AI and reduce staff for some back-office functions by almost 50%, because agents can now perform that work.

So I would go through the P&L and identify which technologies help each cost or revenue line.

As for what I would not do: I have changed my mind somewhat on driver hubs.

For a long time, there was a discussion around autonomous driving where the first mile and last mile were considered complicated, but highway driving was viewed as relatively easier. People considered using driver hubs at different points along interstates.

My guess is that relatively soon, autonomous vehicle technology will be ready to go end-to-end, without needing those driver hubs.

So if you are thinking about building major real estate and facilities along highways, I would be cautious. It may be worth waiting to see whether the technology matures faster than expected.

EV Residual Values and Longer Asset Life

Chris: You mentioned depreciation. EVs have fewer moving parts and may last longer than traditional internal combustion engine vehicles. Do you see an opportunity to hold assets longer and stretch out the value fleet operators get from them?

Philipp: I absolutely believe EVs will hold their value better in the future than they have historically.

So far, EV residual values have been a big problem, especially in the U.S. In the first two years, EVs often saw a much steeper drop in residual value than internal combustion vehicles.

There are several reasons for that. One major reason is that many OEMs reduced prices significantly. If an EV company reduces the price of a new vehicle by $10,000, then the person who bought the same vehicle the day before sees the residual value of that vehicle drop by about $10,000 as well.

We were in a period of price competition, and that affected residual values.

In the future, as the technology matures and as cheaper EVs come to market, I expect this to improve. We already see this in other parts of the world. In China, EVs are being sold at price points comparable to internal combustion vehicles. They have reached price parity and are often total-cost-of-ownership positive over their life cycle.

I think it is only a matter of time before we see more of that in the U.S. as well.

Infrastructure Constraints and Charging Strategy

Chris: We still have a lot of infrastructure challenges to overcome. Some of those challenges have a longer timeline than other technology changes. Will infrastructure happen fast enough?

Dor: In some cases, the problem of getting enough electricity to the right place is still huge.

Many companies are trying to create their own charging infrastructure now to get ahead of competitors and secure the electricity they need, even if they are not using it yet.

Solving this issue will be less about simply providing more power and more about using power more efficiently.

It will be difficult to create new infrastructure quickly. But if you can improve vehicle efficiency or fleet efficiency, the infrastructure problem becomes less significant for your fleet.

Chris: It is hard to know how much companies should invest in infrastructure themselves versus waiting for infrastructure to appear.

Philipp: You are absolutely right. It is a chicken-and-egg problem, and it is particularly pronounced in the U.S. because demand charges create a significant fixed-cost burden for some charging infrastructure.

For fleets, charging infrastructure is easier when vehicles can charge at the depot and are not reliant on public charging infrastructure.

In that case, you can do the math. How many miles do your vehicles drive? What electricity demand do you have? It is relatively predictable. For those fleets, it is easier to justify the investment in charging infrastructure.

Another option is to invest in vehicles with larger batteries, which reduces some infrastructure demand. You might also combine larger batteries with solar, making the fleet less reliant on the public grid.

There are different ways to think about it. The more you rely on another party to build public charging infrastructure outside your depot, the harder and less predictable the transition becomes.

Dor: We have seen examples where companies that want to create infrastructure also create mobility services on top of it. They do that to generate captive demand and guarantee some minimum utilization, helping the unit economics make sense.

Working With New Technology While Keeping Operations Running

Chris: We see a lot of startups and companies bringing new innovations and methodologies to market. We have talked about a bimodal technology approach: trusted existing systems on one side, and experimentation with the future on the other.

Is that an approach you would recommend for fleet managers? How should they interact with new technology and opportunity while keeping the business running?

Dor: I have seen this both from the startup perspective before we joined Element and now from Element as an enterprise interacting with startups and new technologies.

A defining moment for us was when, as a startup, we began working with a fleet operating at JFK Terminal 5. We felt there was a disconnect between the data we were getting from the fleet manager and the actual operation.

So we went there overnight to see the operation ourselves.

We saw a large group of drivers running that operation at the terminal. Many of them did not speak English, but we felt this was the right way to understand how the fleet actually behaved. It was the best source of data.

Instead of relying only on spreadsheets and platform data, we joined the drivers’ WhatsApp group. That gave us direct feedback on how we were affecting the drivers’ day-to-day work.

It is about having the right data, controlling it, and understanding it. But it is also about being in the details and getting the right feedback. Sometimes that feedback comes from computer screens, but sometimes you need to be physically present in the operation to understand how things work and how to improve them.

Chris: I love that. Get out there and see the reality of the operation.

Philipp, what is your view?

Philipp: There is real value in piloting these technologies. Companies need to do that.

But I sometimes wonder whether everyone needs to pilot everything by themselves.

In the future, I think competition will move away from individual companies competing with each other and toward ecosystems competing with other ecosystems.

That ecosystem could include a fleet, an OEM, a telematics player, a fleet management company, and others. Together, they can become a competing ecosystem.

Within that ecosystem, maybe not everyone needs to try everything. Perhaps a few fleets work together and say: you try this technology, I will try that one, and then we compare business cases. Once something works and shows positive P&L impact, both can scale it.

Advice for Fleet Managers in 2026

Chris: If each of you had one piece of advice for fleet managers in 2026, what would it be? Philipp, you start.

Philipp: Stay curious. Try different things. But do not think you can do it all by yourself.

Many of these technologies require moving the entire organization. For new technologies, fleet managers often need to become evangelists within their companies. They need to help convince the organization to change and adopt new capabilities.

Chris: Dor, what is your advice for 2026?

Dor: Keep experimenting, and make sure you have the right partners.

Standing on the shoulders of giants is a great way to start the year. If you have a partner with the right knowledge, aligned technology, and a shared way of seeing the problems you are experiencing, that is a strong way to move forward.

Chris: If people want help from a good partner like Element, where should they go?

Dor: I invite people to visit our new website and learn more about the operations we are doing at Element and how Element can support their operations.

Chris: Philipp, where can people learn more about your work?

Philipp: McKinsey.com is always a good starting point. Some of our knowledge and research comes out under the McKinsey Center for Future Mobility, so I would look for that as well. We are also active on LinkedIn. Anyone listening can find me there, connect, and we can be in touch.

Chris: That sounds great. Thanks so much to both of you for being here. There was a lot of great wisdom shared.

Philipp: Thank you very much.

Dor: Thanks for having us, Chris.

‍