What Legal Innovators Can Learn about Agentic AI from Waymo

Last month, I was in Los Angeles and decided I had to ride in a Waymo. The driverless cars from Google's parent, Alphabet, wander the streets there. They obey traffic signals and stop signs, and even react when a pedestrian overconfidently walks out into an intersection. I set up the app on my iPhone earlier in the day, and after meeting with a client a few blocks from the iconic Santa Monica Pier, I booked a ride back to my daughter's place, where I was staying.

The app directed me to a nearby alley for pick-up. When the car arrived, I opened the door using the Waymo app on my phone and was on my way. The riding experience was smooth and uneventful. After a ten-minute ride, I was out and at my destination. The onboarding and first impression were delightful.

Early Adopters in Los Angeles, San Francisco, and Phoenix may find this experience more routine, but for most people, it is a foreign one.

The Waymo initiative began formally back in 2009, based on earlier research. The first driverless tests on public roads were in 2017, and the commercial service began in 2020.

The neural network AI that drives autonomous vehicles was started years before the emergence of Large Language Models. A number of the leading scientists and AI researchers working on Large Language Models got their start working in autonomous vehicle research. Autonomous vehicles evaluate countless data points and are entrusted with making life-or-death decisions at every intersection. Legal innovators looking to entrust Agentic AI in their firms or law departments can benefit from understanding the successes and challenges of Waymo's autonomous vehicle deployment.

Here are four points from Waymo's progress that can help guide legal innovators as they develop and deploy agentic solutions.

First Impressions Matter

The user experience is critical, especially for first-time users. Waymo goes to great lengths to explain to riders how to use the service, its limitations, and its privacy and safety features. It's okay to experiment, as Waymo did before rolling out its autonomous service, but when it's time to roll out an agentic solution, make sure training, firm policies, and support procedures are clear. This is important even with a controlled rollout. A positive impression is lasting, especially with attorneys. Unfortunately, a bad impression is also a lasting impression. Make sure your first impression is positive.

Methodical rollout and gradual adoption

Waymo is only available in five cities right now. (Austin and Atlanta now have limited service.) With persistence and discipline, deploying an agentic technology practice area by practice area, task by task, or through another prioritization method can lead to successful deployment and avoid surprises. Be careful not to overpromise or overextend.

Expect a little chaos

Waymos do have limitations. They are not allowed to drive on highways in Los Angeles. They struggle in dense traffic and can't pick up or drop off on high-traffic streets. There was the humorous honking incident where Waymos would honk at each other in a parking lot.

Set expectations for agentic applications and be prepared for unintended consequences. Waymos have a steering wheel for a reason. A human can still operate the car. A human can be reached through the app or via the touchscreen in the car if an issue arises. Consider human overrides for agentic workflows and make sure there are governance, guardrails, and human escalation points.

Don't lose sight of the long term

It is said that humans tend to overestimate what they can accomplish in one year but underestimate what they can accomplish in ten years.

Users are very concrete about what technology can do for them right now. They aren't thinking about the incremental improvements that are possible in six months, two years, or ten years.

Waymo plans to expand to Miami, DC, and Dallas in 2026. Waymos can now talk to each other to assist with navigation. They don't honk at each other in a parking lot anymore. Problems like highway driving and pickup location will be addressed, especially when there is a critical mass of "smart cars" navigating roads together. The network effect benefits may be a decade away, but may achieve more than we might imagine.

Legal innovators should deliver results for the here and now. Pleasantly surprise your users with enhancements to agentic solutions that exceed their expectations. As innovators, you can see beyond the here and now to what is possible. Don't expect your users to be thinking about the future. They are busy with deadlines.

Agentic AI is the new buzzword and hype. Legal innovators must cut through the hype to experiment and then deploy solutions that address a well-scoped, narrow problem. Deployment must consider training and easy onboarding. Human oversight and planning for the unexpected are essential. Adoption can be gradual. Expansion to new use cases and greater rollout can be methodical. By managing expectations and delivering incrementally, legal innovators can achieve success similar to Waymo's approach.

When a critical mass of autonomous vehicles talk to each other and work together, more efficient traffic patterns will surface, and passengers should be safer. Similarly, the network effect will drive longer-term benefits for attorneys as more agents are deployed and more attorneys adopt the technology. There will be some chaos along the roadway, and legal innovators should be prepared to manage that.

Deliver for the here and now, but keep looking towards the future!

Image credit: Mliu92, CC BY-SA 3.0 <https://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons Image source: https://commons.wikimedia.org/wiki/File:Waymo_robotaxis,\_San_Francisco_3.jpg