Agentic AI: Is it ready for heavy-duty fleets?

Understanding the distinction between these two AI approaches could be a difference-maker in terms of maximizing efficiency and your competitive advantage, particularly in sectors with complex logistics and freight operational needs.

The difference between Gen AI & Agentic AI

Generative AI excels in creating original content based on patterns and structures identified in training data. This type of AI is especially smart at tasks such as creating text, images, and even code, often mirroring human-like creativity and problem-solving abilities.

Generative AI systems, like GPT-4 or DALL-E, operate by recognizing patterns in vast datasets and using these insights to produce new outputs that resemble the style or content of the original data. However, generative AI typically requires specific prompts or inputs to initiate the creation process and lacks the ability to act autonomously or make independent decisions.

In contrast, agentic AI represents a more advanced and autonomous form of artificial intelligence. Agentic AI systems are designed to continuously analyze real-time data from their environment, make decisions based on predefined objectives, and adapt to changes dynamically. These systems employ advanced technologies such as reinforcement learning and sensor integration to perform tasks independently, without constant human intervention or monitoring.

The key distinction lies in the autonomy and decision-making capabilities of agentic AI, which allow it to act and react in real time to changing conditions and environments.

What does this mean for fleets?

The potential applications of agentic AI are especially compelling, particularly for private fleets. They often deal with complex logistical challenges, including route optimization, fuel efficiency, maintenance scheduling, and real-time decision-making in response to changing road conditions or unexpected events on top of their retail operations, as an example.

And the amount of fleets overall pairing data analytics with some form of AI is increasing. As an example of AI in route planning and optimization, a fleet developed an in-house platform condensing five external platforms. This penta-platform utilized data capture and some algorithms to analyze traffic patterns, weather data, and delivery schedules to determine the most efficient routes for drivers. This has not only reduced fuel consumption and delivery times for their fleet, but also helps in meeting tight deadlines and improving overall productivity.

Furthermore, by analyzing data from telematics, AI-driven predictive maintenance systems can anticipate potential vehicle issues before they occur, reducing downtime and maintenance costs. These systems analyze data from onboard sensors and historical maintenance records to predict when a vehicle is likely to require servicing, allowing for more proactive maintenance scheduling.

AI is also now being used partly for truck procurement, leasing, and financing, in the areas of analyzing market trends, assessing vehicle depreciation rates, and optimizing fleet composition. However, this does not include negotiations with OEMs, finance partners, custom fleet specs, etc. Those components are still best discussed as a team, backed by data-driven decisions.

Today’s organizations are still hesitant to go all-in on the use of AI for procurement decisioning, according to a Fleet Advantage study. Only 19% said they are very confident in this area. And possibly connected, 24% said they are concerned with the accuracy of AI data.

That doesn’t mean all data from AI systems is inaccurate. Fleets can leverage the experience of asset management  partners, who supplement fleet insights with gated, home-grown data from machine learning algorithms to predict the total cost of ownership (TCO). This includes procurement of different vehicle makes, models and types, specs, and helping these companies decide whether to lease or buy, and when to do it.

As an example, they have the ability to constantly scrutinize gated data from learning models and analytics. They pull from an extensive array of data sources, including:

• Vehicle specifications (make, model, year, engine type)
• Operational data (mileage, fuel consumption, route information)
• Maintenance and repair records (repair history, part replacements)
• Financial data (purchase price, interest rates, tariffs, depreciation rates)
• External factors (fuel prices, market conditions, government regulations)

These partners and their analysts then leverage analytics and algorithms to process this data to identify findings that influence TCO:

• Truck spec’s based on safety, fuel efficiency and utilization
• Maintenance and repair frequency and costs
• Depreciation rate and resale values
• Local utilization patterns (e.g., long-haul vs. short-haul routes)

Using this processed data, the models are then trained on gated historical TCO information for various truck models and operational scenarios.

How agentic AI changes the game

The introduction of agentic AI to these processes could significantly reshape the way companies process and utilize data for their fleet operations. Agentic AI has the power to autonomously manage many aspects of fleet operations in real time. For instance, an agentic AI system could continuously monitor and adjust routes based on real-time traffic conditions, weather changes, and even unexpected road closures, making instant decisions to reroute vehicles for optimal efficiency.

According to a recent survey on use of AI in the transportation industry, 95% of companies surveyed said the use of AI for their operations is either important or very important. However, only 19% said they are using the “agentic” form of AI.

For maintenance and vehicle health, agentic AI can leverage data from multiple sources — including onboard sensors, historical maintenance records, and even external factors like road conditions and weather patterns. An agentic AI system could not only predict maintenance needs but also autonomously schedule and coordinate maintenance activities to minimize disruption to operations. This is also important since 62% of survey respondents said they would like to utilize agentic AI for their maintenance operations.

Access to the right data is still everything

However, the effectiveness of any AI system, whether generative or agentic, is fundamentally dependent on the quality and reliability of the data it processes. Today’s leading organizations have quickly realized the value and importance of “gated data”, which refers to high-quality, verified information that has been carefully curated and protected.

The importance of data quality becomes evident when considering the potential consequences of using bad or unreliable data, even leading to hallucinogenic AI outcomes.

For fleets, inaccurate or outdated data could lead to poor decision-making with far-reaching implications. For example, if an organization is leveraging agentic AI with inaccurate fuel consumption data, it might lead to suboptimal route planning, increased fuel costs, and potentially missed delivery deadlines — insight that is also identified by human expertise.

In financial planning and asset management, having computers rely on inaccurate data without any human oversight could lead to misguided procurement decisions, inefficient resource allocation, and a negative impact on the company’s bottom line.

Bad data can also skew financial projections, leading to overinvestment in certain areas while neglecting others.

Organizations with transportation fleets would be wise to investigate the opportunities that both generative and agentic AI can pose for them. However, in the long term, the potential for data-driven mistakes can erode a company’s competitive advantage and financial stability. It is critically important to have access to trusted partners that can continue to oversee the impact these machines have on decision making, as well as access to only the most trusted gated data for extreme accuracy when investing in this technology.