Regarding generative AI, “The ability to train an expert and be able to have that expert reason logically about your data is completely game-changing in my mind,” said Cawse.
Geotab
What happens when you gather over 3,000 fleet industry partners in Las Vegas for Geotab Connect 2024? You dig into connectivity, AI, telematics, data, and technology to understand how they’re all impacting the future of fleet.
You do this by interacting with marketplace partners, resellers, suppliers, and a thousand fleet operators around two show floors, a vehicle ride and drive, networking events, 40 educational sessions, and one giant Sphere.
Held at Resorts World Las Vegas Feb. 14-16, the conference was massive. But Automotive Fleet was able to distill those themes into a half-hour sit down with Neil Cawse, founder and CEO of Geotab. Cawse was joined by Mike Branch, VP of data and analytics, and Sabina Martin, AVP of product management.
Not surprisingly, how fleets and the transportation industry are harnessing artificial intelligence (AI), particularly generative AI, was the central theme of the conversation.
“ChatGPT is at an inflection point,” said Cawse. “The ability to train an expert and have that expert reason logically about your data is completely game-changing in my mind.”
While ChatGPT launched almost as an amusement, will it crunch fleets’ data to produce meaningful outcomes? “We’ve seen the results, and it’s happening,” he said.
Cawse, Branch, and Martin outlined how Geotab is developing AI, how AI is intersecting with numerous fleet processes, and the AI inflection points that will drive fleet innovation.
Next Steps for AI in Fleet Safety
The top use case for telematics in fleets continues to be fleet safety.
Traditionally, telematics data around driver behaviors such as speeding, braking, and harsh cornering has been used to create driver scorecards by assigning a weighted value to these behaviors.
While scorecards have proven effective, assigning these values is an inherently human endeavor that becomes more accurate and predictive when combined with data from millions of Geotab vehicles and context from AI.
“AI is the best way for us to go beyond the traditional risk-based score,” Branch said.
“(Traditional) scorecarding will help drivers improve to a degree, but there’s a point where that will plateau,” Martin said. “This new model will allow us to be more specific and provide more focus areas.”
Cawse (left), Mike Branch, VP of data and analytics, and Sabina Martin, AVP of product management, pause for a photo during a sit down with Automotive Fleet at the 2024 Geotab Connect conference in Las Vegas.
Chris Brown
AI and Autonomous Driving
The path to fully autonomous driving has recently detoured with the slow pace of innovation, drying up of capital, and shuttering of companies. “I do not think that autonomous driving is as close as we all want it to be,” said Cawse.
The hard part is the long tail of reliability, in which engineering autonomous technology from 99.999% reliable to 99.99999% reliable — sufficient for widespread adoption — is 100 times the work.
But while we may never see a completely autonomous world of robotaxis, we’ll see versions of AI in specifically controlled areas such as trucking corridors, Cawse said.
AI still plays a massive role in the development of autonomy, first in its ability to understand the granular context of the world outside the vehicle, from parsing rain from snow and a bouncing ball across a road, and then reacting accordingly.
Geotab is involved in analyzing telematics data to understand how fleets will transition to autonomy. “We understand the movement patterns of these vehicles,” Branch said.
And generative AI will start to play a bigger role.
“The advancements in large language models (LLMs) and generative AI will need five years to feed into what we’re seeing on the autonomous driving side,” Cawse said. “When that happens — and we have lots of people working on this problem — you’re going to see the world change.”
AI and Transparency, Data Security
Regarding AI and data security, Branch pointed out that Geotab customer data is not used to train LLMs or tools like ChatGPT.
But while the data itself always stays within the restricted Geotab environment a question prompt disassociated from data can inform an LLM. These distinctions are important, and Geotab is preparing documentation to address issues around data security, transparency, and open dialogue.
“Responsible AI is absolutely at the top of things we think about,” Cawse said.
Geotab is also working to understand how AI can be abused and what prompts cross the line. “For instance, the LLM shouldn’t answer a question like, ‘Should I fire my fleet driver?’” Branch said.
AI will never get it right all the time, Branch said, so it’s crucial to create checks and balances.
Employing AI Agents
While the last 18 months have seen a groundswell of activity around LLMs and generative AI, Cawse made the point that Geotab has been using AI in various other forms for 10 years.
This includes anomaly detection, such as identifying unsafe driving maneuvers or detecting collision damage, and more recently understanding the health of an EV battery and when to replace it.
A bigger inflection point, but more in the future, will be around agent-based infrastructures, in which users will converse with AI “agents” to solve tasks in areas such as safety or maintenance.
Where humans spend hours on these tasks manually today, AI can automate and perform them infinitely quicker. “Those admin-heavy tasks, like compliance tasks, are a real opportunity for those agents as well,” Martin said.
Branch cautioned that work must be done to improve trust in the system and ensure that it’s doing the right thing.
AI and Redundancy
Technology implementation and automation have traditionally and will in the future cause workforce realignments.
Regarding AI, “We don’t see it as being a replacement for people’s jobs,” Cawse said, or at least it will be many years before this bridge will need to be crossed.
AI should be viewed as a powerful copilot that is like an assistant that works side by side with you, he said.
“AI will just make your life a lot easier. But at the end of the day, you’re the one deciding what gets done, how it gets done, and when it gets done. You’re in control.”
