Traditionally, the transportation industry has taken a cautious approach to addressing business challenges and adopting new technologies. Because assets are expensive, margins are consistently thin and mistakes can have massive real-world consequences, fleets simply cannot afford to be early adopters of unproven solutions.

Yet, when it comes to artificial intelligence (AI), our industry has set aside this pragmatic approach and entered a new period of rapid experimentation. Many fleets today are deploying AI-powered maintenance tools, using smarter TMS solutions and automating back-office workflows.

The opportunity this presents is real. But so is the risk. When a contract is executed incorrectly or a carrier is wrongly flagged, someone must explain what happened and why. In a multi-agent system, the audit trail can become confusing, making it difficult to pinpoint where the breakdown in communication occurred.   A study by Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027 due to governance and accountability complexities.

As AI becomes embedded throughout transport operations, many organisations are unintentionally creating a new form of fragmentation that poses serious security risks to the supply chain. Instead of deploying one secure AI agent that can automatically use data to connect all workflows and operations, organisations are accumulating dozens of disconnected AI tools that work independently across different data sources and security standards.

The result is a tech environment that may be more advanced on paper, but not necessarily more efficient, secure or manageable in practice.

Rapid AI adoption compounds security risks

Today’s off-the-shelf consumer-grade AI tools are inexpensive and easy to access, but most are unfit for logistics operations at scale.

Many of these AI tools lack effective security oversight and often work around security controls in an attempt to complete requested tasks. True security demands those boundaries to be enforced without compromise.  In extreme cases, AI agents have installed unauthorised software on host machines, creating new API connections and other unapproved actions.

The “slap it on” AI strategy that many companies are using to try to move faster is opening a Pandora’s box of avoidable risks and security threats.  Transport and logistics leaders should develop a strategic AI rollout plan that takes security risks seriously without moving too slowly and getting left behind. Finding the right balance is easier said than done.

Security must be baked into one single AI agent

One thing is clear: if businesses are to continue down the path of safely embracing agentic AI, fleets need to rethink the mix of AI tools that they have been leveraging for their business operations.

In order for a single AI agent to be successful, security must be baked into the very foundation of its development, with a standalone team focused on security as its main objective. Additionally, numerous overlapping layers of defence must be put in place to prevent both intentional and negligent use in real time, and bring a human into the loop whenever tasks are unclear or run into security restrictions.

Replacing 12 separate agents (and audit trails) with one eliminates much of the governance and accountability complexity organisations experience today.

Fleets today would benefit from a single, powerful AI agent that is trusted to operate with access to live business data, internal servers, strict cost controls and strong user-permission constraints.

Fragmented AI reverses past efficiency and security gains

On the surface, it may appear that fleets who rapidly adopt new AI tools and throw them at gaps in their operations are getting ahead. But the reality is, despite this growing adoption, many fleets are left with fragmented systems and exposure to new security threats.

When AI is not deployed strategically across a business the efficiency gains these tools should deliver can begin to unravel, creating inefficiencies instead.

In a survey of 68 freight professionals, about 55% of respondents said they were actively using, piloting or experimenting with AI in limited areas, but only about 15% were using AI agents.

With so much potential for efficiency, our industry should move towards seamlessly integrating a singular AI agent that touches all business aspects including telematics, third-party software and automated workflows to enable transport operations to move forward.