At a carrier in Nashville, a 22-year veteran driver named Carl hears from his daughter that "AI is replacing truck drivers." He has been hearing versions of this story for three years. He is not sure whether to believe it, but he is sure that the new system the dispatcher uses to assign loads has something to do with AI, and he is pretty sure it's the reason he got the Chicago run last Tuesday when he was supposed to be home by Thursday. His unease is not irrational and his skepticism is not ignorance. He is asking the right question badly: not "will AI take my job" but "does the organization that runs me on AI understand what it is doing with my livelihood?" The answer, at most carriers in 2026, is incomplete. The technology is ahead of the explanation, and the explanation is ahead of the training, and nobody has thought through what Carl needs to know and why his knowing it matters to the carrier as much as it matters to him.

Building an AI-literate fleet workforce is not a training initiative. It is a capability-building program that starts with the driver who has to trust that an AI-assisted dispatch decision respected their HOS (hours of service) clock and their home-time promise, runs through the dispatcher who must understand what the optimization model is doing well enough to catch it when it is wrong, extends to the shop manager who needs to know why a predictive alert fires and what the shop's job is when it does, and reaches all the way to the director and the owner who must make investment and governance decisions about AI programs they cannot interrogate if they do not understand the basics. This is not one audience. It is five, and each needs a different curriculum.

The carriers that are winning in 2026 against the structural challenges of an 80,000-driver shortage and an autonomous transition that is already live in their TMS (transportation management system) are not winning because they bought better AI tools. They are winning because they built the organizational capacity to use those tools well. Their dispatchers understand what the model optimizes and where it fails. Their drivers understand what AI does and does not do to their loads. Their shop managers understand why the alert fired before they decide whether to pull the truck. Their Fleet AI Lead can interrogate a vendor's model claims without being deceived. Their directors understand the accountability structure well enough to defend AI program decisions to an owner, a shipper, or a Federal Motor Carrier Safety Administration (FMCSA) auditor. Literacy at every level is the multiplier that turns a technology investment into a competitive capability.

Why Fleet AI Literacy is a Safety and Retention Issue, Not Just a Training One

Framing AI literacy as a training initiative makes it easy to underfund and easy to forget after the vendor's onboarding sessions are done. The reframe that matters is this: AI literacy in a fleet is a safety and retention issue, and it has the same urgency as HOS compliance training or DVIR (driver vehicle inspection report) completion.

On the safety side, the connection is direct. When a shop manager acts on a predictive-maintenance alert without understanding what the alert is telling them, they make the wrong call more often. When a dispatcher approves an AI-generated match without understanding the model's HOS logic, they are more likely to miss a violation the model introduced. When a driver does not understand that their ELD (electronic logging device) data feeds the dispatch model that proposes their next load, they cannot flag an ELD error that is generating wrong HOS availability signals and causing the system to overload them. Literacy at each role is not abstract. It is the specific understanding that makes the human review step in the AI-enabled operating model actually work, and without it, the carrier is running a governance structure that exists on paper but not in practice.

On the retention side, the connection is equally direct. In a freight market with an 80,000-driver shortage and 237,600 annual openings projected through 2034, the average driver age is 46 to 47, and the industry is fighting for every experienced driver. Carl's anxiety about AI is not going away because nobody addresses it. It either gets resolved by a carrier that takes the time to explain what AI does in his lane, or it motivates him to find a carrier where someone at least pretended to care about the question. The cost of losing an experienced driver is not just recruitment and onboarding. It is the loss of lane knowledge, shipper relationship knowledge, and operational reliability that no new hire brings on day one. In a shortage this severe, literacy that keeps experienced drivers engaged is a retention program, and it costs a fraction of driver replacement.

The case for AI literacy also extends upward. Directors and operations VPs who do not understand AI governance basics cannot ask meaningful questions about program decisions, cannot evaluate vendor proposals without being deceived, and cannot defend AI-related safety or compliance decisions to a regulator without a literacy foundation. At the board level, an owner who does not understand the basics of what the fleet's AI program is doing cannot exercise the oversight the carrier's own governance policy and FMCSA's evolving driverless-operations rules expect. Literacy at the top of the org chart is not about understanding model architecture. It is about being able to ask and understand the answers to three questions: Is the AI working? Is it safe? Is it legal?

AI literacy in a fleet is a safety and retention issue with the urgency of HOS compliance training: without it, the human review steps that protect the carrier exist on paper but not in practice.

The Driver Curriculum: From Threat to Tool

The driver audience is the largest in the carrier workforce and the most emotionally charged when it comes to AI. The curriculum for drivers has to start where Carl is: with his actual question, which is not "how does machine learning work" but "will this thing make my work life worse, and who decided that was okay?" Addressing the technical question without addressing the human question first produces a training session drivers sit through and promptly forget.

The first module for drivers is called "What AI Does and Does Not Do to Your Load." It has three learning objectives. First, drivers learn that AI assists the dispatcher in proposing load matches, but that no AI system automatically commits a driver to a load at a compliant carrier: the dispatcher makes that commitment and owns it. This is not a philosophical point. It is a practical one: if Carl calls the dispatcher to say the Chicago run doesn't work for his home-time, he is talking to a human who has the authority and accountability to respond, not to a software decision. Second, drivers learn what data from their ELD feeds the dispatch model and why accuracy in that data matters to them: an ELD that is not synced correctly shows the model wrong HOS availability, and the model may propose a load that the driver cannot legally run, which creates the exact situation Carl experienced in Nashville. Third, drivers learn the channel for flagging what feels wrong: if a load assignment feels like it violated a home-time commitment or an HOS limit, there is a dispatcher and a process, and the driver's voice in that process is both their right and a quality signal the carrier needs.

The second module addresses autonomous trucking directly, because the rumor mill about "driverless trucks replacing everyone" is already running in every truck stop in the country. This module uses the real numbers and the real operating model. Aurora has logged more than 250,000 commercial driverless miles and is bookable through McLeod TMS integration, but it operates on limited interstate highway lanes between transfer hubs. The driver who takes the load from the origin to the transfer hub and from the transfer hub to the final-mile destination is not replaceable by current autonomous technology. The driver who handles a shipper with a difficult loading dock, weather that the autonomous system's operational design domain does not cover, or a customs delay at a border crossing is not replaceable. The lesson is not that drivers are safe forever. It is that the current and near-term autonomous transition creates new driver roles rather than eliminating the need for human freight expertise, and that the driver who understands those roles and positions for them is far better placed than the driver who simply fears the transition.

The third module covers the fairness and accuracy of AI-assisted driver scoring, because many carriers have already deployed AI safety-scoring and coaching systems that rate drivers on hard-braking events, speed, and other telematics metrics. Drivers who do not understand how these systems work cannot advocate for themselves when a score seems wrong, cannot understand what to change in their driving to improve their score legitimately, and cannot identify when a score is reflecting a sensor error rather than a driving behavior. This module teaches drivers to read their own AI-generated safety score, to understand what events it is based on, to ask a specific question when a score seems wrong (which event, which date, which unit), and to know who at the carrier owns the review of disputed scores. The module also teaches carriers: a scoring system that drivers cannot understand or contest is a retention problem waiting to happen, and in a shortage market it will happen.

The Dispatcher Curriculum: Understanding the Tool You Depend On

The dispatcher audience needs the deepest technical literacy of any frontline role in the carrier, because the dispatcher is the human whose review is the primary safeguard in the AI-enabled dispatch operating model. A dispatcher who does not understand what the AI optimization model is doing cannot review its proposals meaningfully, which means the safeguard does not actually exist despite appearing on the org chart.

The dispatcher curriculum has four modules. The first is "What the Optimization Model Is Doing": the dispatcher learns that the model is solving a constraint-satisfaction problem with the carrier's objective function (minimize deadhead miles, maximize revenue per mile, respect HOS and home-time constraints) against the current state of the dispatch board (available drivers, available loads, equipment types, lane restrictions). The dispatcher does not need to understand the mathematics. They need to understand the objective function, because the objective function determines what the model cares about and what it does not care about. A model that is told to minimize deadhead and maximize revenue per mile will generate proposals that score well on those dimensions, and it will not spontaneously consider the driver who had a hard week, the shipper who needs special handling, or the lane that has a reliability pattern the model's training data does not capture. The dispatcher's review is the mechanism that introduces those considerations into the decision.

The second module is "Where the Model Makes Mistakes." This is the module most vendors skip in their onboarding because they do not want to train users to doubt their product. But a dispatcher who knows the model's specific failure modes catches them. The three most common failure modes in dispatch optimization models are: stale HOS data (if the ELD sync is delayed, the model works from incorrect HOS availability), missing lane constraints (the model does not know about a shipper's informal but enforced detention pattern, a road closure that is not in the mapping data, or a lane restriction the carrier added after the model was configured), and objective function edge cases (the model finds a technically legal and marginally optimal match that is practically wrong for reasons the objective function does not capture). Dispatchers who know these failure modes look for them in every proposal they review. Dispatchers who do not know them miss them.

The third module is "How to Override and Why It Matters." Dispatchers who are uncomfortable overriding AI proposals are a governance problem. Override hesitation leads to rubber-stamping, which eliminates the human accountability layer the carrier's operating model requires. This module teaches dispatchers that overriding a proposal is not a failure of the AI system or a reflection on the dispatcher's ability to work with AI. It is the human judgment step the operating model requires. The module uses real examples of legitimate overrides (driver context the model cannot see, shipper relationship context, a lane reliability problem that post-dates the model's training data) and teaches dispatchers to document overrides with brief reason codes that feed the Fleet AI Lead's model quality monitoring. An override is valuable data. A rubber-stamp approval of a wrong proposal is a compliance and liability risk.

The fourth module is "The HOS Gate and Why You Own It." This module covers the dispatcher's non-delegable responsibility to verify that every dispatched load is HOS-legal regardless of what the AI model proposed. It covers what the common HOS violation looks like in the model's output (a load that assumes hours the driver will not have after a planned stop, a calculation that does not account for a required 30-minute break), how to catch it in the AI-generated HOS summary before committing the load, and what the consequences of missing it are: an FMCSA violation, a CSA (Compliance, Safety, Accountability) score hit, and a driver forced to go out of service on the road. The dispatcher who owns this module is the carrier's primary compliance safeguard at the dispatch step, and they need to know it.

The Shop Manager and Tech Curriculum: Understanding Alerts and Owning the Call

The shop audience has two distinct sub-audiences: the shop manager, who owns the judgment calls on predictive alerts and return-to-service decisions, and the technician, who does the physical work and owns the accuracy of the repair documentation. Their curricula overlap but are not the same.

For the shop manager, the core curriculum has three modules. The first is "What the Predictive Alert Is Telling You." The predictive maintenance AI is applying statistical pattern recognition to telematics data: it is saying that the combination of signals it is observing in this vehicle resembles the signal pattern that historically preceded failures of this component type within a certain time horizon, at a certain probability. It is not saying the component will fail. It is not saying the component is currently defective. It is saying the risk profile of this vehicle has changed, and the change is worth a human's attention. The shop manager who understands this can ask the right follow-up questions: what is the historical failure rate associated with this alert type on this platform, what is the failure-to-alert accuracy of this particular fault pattern, and what is the in-shop inspection versus the roadside breakdown cost comparison for this component? This is a materially better decision than "the AI flagged it, so I'll pull it" (over-reaction) or "the AI flags everything, so I'll wait for Thursday's PM" (under-reaction).

The second module for the shop manager is "Documenting the Decision You Made." Every time a predictive alert is reviewed and acted upon or deferred, that decision needs to be documented in the work-order system. This documentation is not bureaucratic overhead. It is the carrier's defense when a truck that had a prior alert subsequently breaks down on the highway. The shop manager who can produce a documented review note showing "Alert reviewed on [date], inspection performed, component within spec, cleared for service" is in a fundamentally different legal and regulatory position than the shop manager who cannot demonstrate that the alert was reviewed at all. This module teaches the minimum documentation content, the work-order system fields that capture it, and the review cycle in which the Fleet AI Lead will audit it.

The third module for shop managers is "Working with the AI's Suggested Work-Order Sequence." The AI-generated work-order prioritization is a recommended sequence, not an order. The shop manager has authority and accountability to modify it. This module uses three scenarios where the right modification is obvious (a tech with specific skills makes a different sequencing better, a part that arrives this afternoon changes the logic, a driver needs their truck early for a family commitment) and three where the temptation to modify without a good reason is real but the AI's sequence is better. The goal is to build the shop manager's confidence in interacting with the AI as a tool, not as an authority.

For technicians, the curriculum is simpler: one module on "Using AI-Assisted Documentation Without Losing Ownership." The tech learns that AI drafts the work order from their verbal notes or structured inputs, that their review of the draft is not optional, that correcting errors is part of the job even when the draft is usually accurate, and that their signature on the work order is the transfer of accountability from the AI draft to the human record. The tech also learns why this matters beyond the carrier's internal quality: incorrect work-order documentation can void warranty claims, can appear in FMCSA maintenance audits, and can be introduced in legal proceedings after an accident. This is not a lecture on legal liability. It is a clear statement that the tech's name on the document means something, and that something is that they verified it is accurate.

The Director and Owner Curriculum: Governance Without Technical Depth

The director and owner audience does not need to understand how a neural network works or how a predictive model is trained. They need to understand three things: what decisions AI is making in the organization, who is accountable for verifying those decisions, and how the carrier knows whether the AI program is working, safe, and legal. Everything in the director and owner curriculum serves those three questions.

The first module is "Reading the Fleet AI Performance Report." The director and owner learn to read the quarterly AI governance report the Fleet AI Lead produces: what deadhead trend means and why it matters, what breakdown rate trend means and why it matters, what the CSA score trend connected to AI-monitored driver behavior means, and what autonomous lane economics mean for the carrier's long-term competitive position. They also learn what the process metrics mean: an AI proposal acceptance rate that is too low signals a model quality problem or a training problem; an override rate concentrated on a specific lane signals a model input gap; an alert review rate below 95% signals an operating model failure. Directors and owners who can read this report can ask the right questions at the quarterly review. Directors and owners who cannot read it either skip the review or defer to whoever is presenting it, which means the governance oversight function is ceremonial rather than substantive.

The second module is "AI Vendor Evaluation: The Questions That Protect the Carrier." This module is not a technical deep dive. It is a set of questions that any director or owner can ask an AI vendor before signing a contract, and a clear statement of what acceptable and unacceptable answers look like. Can you show us the performance of this model on a fleet profile similar to ours? Can you document the fair-treatment provisions in your model design for driver-facing AI? What are the contractual audit rights this agreement gives us if we need to investigate an AI-related incident? What is the model's update process, and how will we be notified of changes that affect our operations? What data generated by our fleet remains ours if we terminate the contract? Directors and owners who can ask these questions and evaluate the answers protect the carrier from the vendor relationships that most often create governance and liability problems: the ones where the carrier signed a contract without understanding what it was agreeing to.

The third module is "The 90-Day AI Governance Calendar." This is a practical module that gives directors and owners a repeatable structure for staying engaged with the AI program without requiring deep technical involvement. The structure is: monthly brief from the Fleet AI Lead covering any incidents, any process-metric flags, and any decisions that need leadership input; quarterly governance report with the full metric set and the forward roadmap; and a biannual vendor review that assesses whether current tools are performing against the original business case and whether the market has produced better alternatives. The director who shows up to these reviews with the performance report already read, three questions prepared, and a standing expectation that the Fleet AI Lead will bring problems up rather than manage them quietly is performing the governance function the carrier's safety and the carrier's owner both require.

Building the Literacy Program: Structure and Delivery

A literacy program that exists as a set of PowerPoint slides shared in an all-hands meeting is not a literacy program. It is a communication. The difference is practice, feedback, and an organizational expectation that the capability is required, not optional.

The program structure at the carrier level should have five elements. First, a role-based curriculum map: each role (driver, dispatcher, shop manager/tech, director/owner, Fleet AI Lead) has a defined curriculum with defined learning objectives, defined assessment criteria, and a defined refresh cycle. The curriculum map is owned by the Fleet AI Lead and reviewed annually or when a significant AI tool or regulatory change occurs.

Second, a delivery model that fits the audience. Driver literacy does not get delivered in a conference room through a two-hour slide deck. It gets delivered in small group sessions at the terminal, in 15-to-20-minute modules that fit between runs, with drivers who are already AI champions (drivers who have had a good AI-assisted experience and are willing to share it) as the faces of the content rather than management. Dispatcher literacy can be delivered in workshop format because dispatchers are office-based, but it needs practice time with the actual AI tools, not theoretical discussion. Shop manager and tech literacy can be delivered in the shop, using real telematics alerts from the carrier's own fleet as training examples. Director and owner literacy is most effective as a quarterly structured conversation rather than a formal training event.

Third, a baseline assessment and gap analysis before the program launches. The carrier that launches AI literacy training without knowing what its people currently understand is guessing. A short baseline assessment (it can be as informal as ten questions for drivers and fifteen for dispatchers) tells the Fleet AI Lead where the gaps are, which roles need the most intensive curriculum, and which misconceptions are most common. At many carriers, the most urgent misconception is not "AI will take my job" (which gets the most airtime) but "the AI must be right or the dispatcher would have overridden it" (which is the misconception that creates compliance risk when it takes hold in driver culture).

Fourth, AI champions at every level. The literacy program should identify, early, the individuals at each level who are most naturally engaged with AI tools and most trusted by their peers, and invest in making those individuals the visible faces of the program. At the driver level, this is the driver who figured out that the new dispatch system accounts for their home-time preference better than the old phone tree and is willing to tell that story at the terminal meeting. At the dispatcher level, this is the dispatcher who caught a model input error and fixed it by flagging it to the Fleet AI Lead, and who is willing to walk new dispatchers through what the model is actually doing. At the shop level, this is the tech who used the AI-drafted work order and found it saved 20 minutes per repair, and who tells the other techs the review step is worth it. Champions do not replace the formal curriculum. They make the curriculum credible.

Fifth, a feedback loop that connects literacy program outcomes to AI program governance. The literacy program should generate observable signals: dispatcher override rate (too high or too low relative to model quality expectations, both indicate training gaps), shop manager alert review rate (below 95% indicates a coverage or motivation problem), driver satisfaction with AI-assisted dispatch (measured in the existing driver satisfaction survey). These signals go to the Fleet AI Lead's quarterly governance report alongside the operational metrics. If the literacy program is working, the process metrics should reflect it. If the process metrics are flagging problems that look like literacy gaps, the Fleet AI Lead adjusts the curriculum before adjusting the AI tool.

The Long Game: Literacy as a Competitive Advantage

There is a short game and a long game in fleet AI. The short game is buying a dispatch-optimization tool and measuring the deadhead improvement. The long game is building an organization that can use AI well across every role, improve continuously as the tools evolve, adapt to the autonomous transition without structural disruption, and maintain the trust of the drivers and dispatchers who run the operation every day.

The carriers that are playing the long game in 2026 have already noticed a competitive pattern: the AI program's value compounds over time for organizations that have built the literacy to use it well, and declines for organizations that treat AI as a tool that works automatically once installed. The compounding happens because a dispatcher who understands the model catches its errors and flags them for improvement, and the model improves. A shop manager who understands the alert makes better pull-from-service decisions, and the shop's maintenance cost curve falls below the carrier's competitors. A director who understands the quarterly AI governance report makes better vendor and investment decisions, and the fleet's AI capability advances ahead of the market.

The decline happens because a dispatcher who doesn't understand the model rubber-stamps proposals until an HOS violation or a missed home-time promise erodes driver trust, and the carrier starts losing drivers in a shortage market where it cannot afford to lose them. A shop manager who doesn't understand alerts either over-responds (pulling trucks that didn't need it, burning bay time and driver goodwill) or under-responds (clearing alerts that should have been acted on until a roadside breakdown puts the carrier's safety record in an audit). A director who doesn't understand the AI governance report cannot catch a vendor whose model is drifting, a Fleet AI Lead who is burying problems, or a program that is costing more than it is producing. The literacy gap is the place where a freight AI investment quietly stops working.

The autonomous transition raises the stakes further. As Aurora's platform and its competitors extend their operational design domains, as FMCSA updates its driverless-operations rules, and as the autonomous long-haul market grows from $2.7 billion in 2024 toward $42.6 billion by 2034, the carrier workforce will face an accelerating rate of change in what the tools do and what the humans are for. The carrier that has built AI literacy at every level -- drivers who understand their evolving roles, dispatchers who know how to manage a mixed human-autonomous board, shop managers who understand how telematics data from an autonomous unit differs from a driven unit, directors who can read an autonomous lane economics report critically -- is the carrier that navigates that transition as a capability-building process rather than as an emergency. The carrier that has not built that literacy is the carrier whose people learn about autonomous expansion by reading about it in a trade publication after it is already happening to them.

Key Takeaways

• Fleet AI literacy is a safety and retention issue with the urgency of HOS compliance training: without it, the human review steps that protect the carrier against FMCSA violations and driver-trust erosion exist on paper but not in practice.
• The driver curriculum addresses the emotional question first (will AI make my work life worse, and who decided that) before the technical one, and covers three topics: what AI does and does not do to load assignment, the real role of human drivers in the autonomous transition (transfer hubs, first-mile/last-mile, domains the autonomous operating design does not cover), and how to read and contest an AI-generated safety score.
• The dispatcher curriculum is the deepest at the frontline level and covers four topics: what the optimization model's objective function is and what it does not consider, the three most common dispatch model failure modes (stale HOS data, missing lane constraints, objective function edge cases), how to override meaningfully and document the reason, and the non-delegable HOS gate responsibility.
• The shop manager curriculum covers three topics: what the predictive alert is statistically telling them and how to ask the right questions before acting, how to document the pull-from-service or deferral decision that is the carrier's defense if a subsequent incident is investigated, and how to use the AI's recommended work-order sequence as a tool rather than an order.
• The director and owner curriculum covers three topics: reading the quarterly AI performance and governance report, asking the vendor evaluation questions that protect the carrier from bad contracts and ungoverned deployments, and following a 90-day AI governance calendar that keeps leadership engaged without requiring deep technical involvement.
• The program structure requires five elements: a role-based curriculum map with defined objectives and refresh cycles, a delivery model that fits each audience (small group for drivers, workshop for dispatchers, shop-based for techs, structured conversation for directors), a baseline assessment and gap analysis before launch, AI champions at every level, and a feedback loop that connects literacy program outcomes to the Fleet AI Lead's quarterly governance report.
• The competitive advantage of AI literacy compounds over time: the dispatcher who understands the model catches its errors and improves it, the shop manager who understands alerts makes better decisions and reduces the carrier's maintenance cost curve, and the director who understands the governance report makes better investment decisions. The literacy gap compounds the other way, quietly degrading a program until a violation, a breakdown, or a driver departure makes the cost visible.
• The autonomous transition raises the stakes: as Aurora's platform and its competitors expand, the carrier workforce will face an accelerating rate of change in what the tools do and what humans are for, and the carrier with literacy built at every level navigates this as a capability-building process rather than as a workforce crisis.