EpochX logistics analysis and worked examples

Reviewed 28 Sep 2026 · Analytical reconstruction and illustrative calculations

Saved research evidence used by the youtube deck. This is a dated provider snapshot or an EpochX analysis, not a live market feed.

# Inside Hertz's AI logistics

Research reviewed 28 September 2026. This memo accompanies slides 10-26 of the 70-slide V9 deck. It distinguishes company demonstrations, generic platform capabilities and our own operating analysis.

## The investment mechanism

Our hypothesis is that joined operating records let Hertz choose and execute better actions on its physical fleet. The economic pathway is specific: a changed decision must alter paid cost, service quality, profitable vehicle availability or capital required. Count completed work and realized contribution, then subtract the cost of software, integration and operating change. Application usage alone does not establish return.

## VIN: identity, relationships and time

A VIN identifies a vehicle; it is not itself a location sensor. In our reconstruction, a vehicle record links to a rental unit, reservation, inspection, work order and transport task. Each linked record needs its own identity, timestamp and owner. A feed may arrive late, duplicate an event or refer to an outdated rental-unit identifier. Joining the wrong records could clear a hold on the wrong car.

Three practical design questions follow. First, which record controls whether the car is rentable? Second, how does a new inspection override an older availability record? Third, does the reservation system receive and acknowledge the change? Hertz has not published its full schema, refresh intervals, identifier reconciliation or interface contract. These are engineering questions to investigate, not claims that the company uses a particular implementation.

## One vehicle, several dependent decisions

Slides 12-13 follow Hertz's company demonstration. Its importance is the dependency between functions. A repair decision changes the time at which a car can fulfill a booking. A replacement assignment may require another move. The move consumes a driver's available time, and the destination needs the right capacity when the car arrives. Optimizing one department's invoice in isolation can make the total journey more expensive.

Our analytical cost rule for a feasible repair option is: repair invoice + transport cost + expected lost rental contribution + expected rework or exception cost. Warranty recovery belongs in the cash comparison; uncertain recovery should not be treated as certain cash. Feasibility comes first: safety, parts, skills and capacity must permit the work. The actual Hertz objective function is undisclosed.

An illustrative example on slide 14 compares $100 plus four unavailable days at $40/day contribution with $200 plus one unavailable day. Totals are $260 and $240. Spending another $100 reduces total economic cost by $20 under those assumptions. The break-even lost contribution is $100/3 = $33.33/day. With weak demand, the faster option may have little incremental revenue value. This is our model, not a measured Hertz result.

## Employees: translate a plan into physical work

The public demonstrations show employees interacting with fleet tasks; the deck separately cites management's labor-planning comments. Our diligence focuses on execution: a task must reach an available worker, be accepted, have keys and access available, and end in a verified physical outcome. A beautifully optimized queue is useless if staff cannot perform the first job.

Measure task acceptance latency, reassignment frequency, waiting time, completed jobs per paid hour and exception causes. Distinguish actual paid savings from a manager's ability to redeploy hours to better service. Shorter task time may improve service without reducing payroll. A staffing recommendation also needs local working-time rules, skills, breaks and minimum coverage; their exact Hertz implementation is not public. Public sources reviewed do not establish automated employee rankings or individual hiring and firing decisions.

## Why hourly planning matters

Slide 20 uses assumed values: 240 vehicle returns over eight hours at 15 minutes of work each create 60 work-hours. At 75% productive time, each eight-hour shift supplies six work-hours, giving ten shift-equivalents across the full period. If 120 of those returns arrive in two hours, the peak needs 30 work-hours. Each person supplies 1.5 productive hours in that window, requiring 20 people then to avoid a growing queue. Overlapping shifts or redeployment may meet the peak. These are coverage requirements, not a recommendation to hire 20 full-day staff. Task times, arrival patterns and the 75% assumption are illustrative.

Slide 21 applies a steady-state flow relationship. At 600 vehicles/day and six hours average turnaround, 600 x 6/24 = 150 cars are in the process on average. At four hours, 100 cars are in the process. The 50-car difference is capacity available sooner, with unchanged throughput. It is not automatically 50 extra rentals per day. Economic value depends on customer demand and deployment; do not count both a full growth benefit and a full fleet-reduction benefit from the same capacity.

## Separate AI, optimization, rules and people

AI interpretation can organize ambiguous observations. Forecasts estimate demand or completion times. A constrained optimizer compares feasible allocations. Deterministic rules can prevent an invalid status change. People handle exceptions and execute physical work. A useful error report separates these stages: misclassified fault, inaccurate time forecast, infeasible move, rejected action or incomplete task. Hertz names both Foundry and Navidea in fleet planning; assigning all improvement to a language model would obscure the mechanism.

## From a recommendation to a durable update

Palantir's documentation describes objects and links, action types, submission criteria and notifications or external-system calls. These capabilities do not establish which interfaces Hertz installed. Our proposed implementation test is end-to-end: initiate a permitted change, confirm it in the destination system, complete the physical job, and check that the next department sees the outcome. Retry handling should avoid duplicate orders; a failed write should enter a visible exception queue. Record who or what approved the action and which input state supported it. These are recommended controls, not a disclosure of Hertz's internal security rules.

## The military comparison

The appropriate analogy is operating readiness. An equipment inventory alone cannot tell a commander what is available for a task. Likewise, an owned car cannot earn a rental while the necessary repair, driver or assignment remains unresolved. The public Army Vantage record describes integrated information supporting operating decisions. An Army predictive-logistics solicitation also illustrates that getting timely vehicle data to maintainers can itself be unfinished development work. A solicitation is not evidence of a completed deployment or achieved savings.

Our inference is that the transferable discipline is a shared operating picture joined to executable work. Military and rental environments have different objectives, data, constraints and success measures. We do not import a military savings figure, mission result or classified capability into the Hertz model.

## What would persuade us financially

For each deployment cohort, capture the previous process, recommendation, human response and realized outcome. Compare similar operations over the same season with controls for demand, fleet age, mix, recalls and local wage changes. Measure customer fulfillment and service outcomes as well as cost. Reconcile recovered availability to incremental contribution or lower capital needs. Reconcile labor gains to paid cost or explicitly valued service improvements. Deduct implementation, licenses, cloud use, support, training and continuing data maintenance. A credible result needs the same scope and period on both sides of the calculation.

The case is credible as an operating mechanism. Public disclosures reviewed do not yet isolate a net financial return caused by Palantir. That distinction is the question the research should resolve over time.

## Source trail

- [Hertz Daytona demonstration at AIPCon 10](https://www.linkedin.com/posts/palantir-technologies_at-aipcon-10-hertz-reveals-how-palantir-activity-7470120635973799936-nqem)
- [Hertz CIO at AIPCon 7](https://www.linkedin.com/posts/palantir-technologies_from-two-way-radios-and-dispatchers-to-custom-activity-7338567961638547457-rf4a)
- [Hertz confirms Daytona deployment](https://www.linkedin.com/posts/hertz_navigating-the-operational-complexity-of-activity-7470127003765985281-u_-A)
- [Original Hertz and Palantir partnership](https://newsroom.hertz.com/press-releases/press-release-details/hertz-partners-with-palantir-to-drive-operational-excellence-and-enhance-customer-experien/)
- [Hertz Q1 2026 earnings call, company transcript](https://s204.q4cdn.com/384814028/files/doc_financials/2026/q1/Q1-Earnings-Call-Transcript.pdf)
- [Q2 2026 management prepared remarks](https://newsroom.hertz.com/articles/article-details/hertz-global-holdings-inc-q2-2026-prepared-remarks/)
- [Palantir Ontology overview](https://www.palantir.com/docs/foundry/ontology/overview)
- [Palantir action types](https://www.palantir.com/docs/foundry/action-types/overview)
- [Palantir action submission criteria](https://www.palantir.com/docs/foundry/action-types/submission-criteria)
- [Palantir notifications and external-system writes](https://www.palantir.com/docs/foundry/action-types/side-effects-overview)
- [Palantir AIP architecture and model tools](https://www.palantir.com/docs/foundry/aip/overview)
- [U.S. Army CIO: Army Vantage](https://www.army.mil/cio)
- [Palantir Army Vantage deployment](https://www.palantir.com/army-vantage)
- [Army solicitation for predictive-logistics data transfer](https://armysbir.army.mil/topics/secure-wireless-data-transfer-solutions-predictive-logistics/)
- [The Army’s Vantage Point: field adoption](https://www.army.mil/article/237749/the_armys_vantage_point)