Begin with a delayed decision, a trustworthy record and someone responsible for acting.
EXECUTIVE ANALYSIS | H. Max Ottesen
AI-generated conceptual editorial illustration. Fictional people and setting; not documentary evidence. Illustrated warehouse handoff with staged goods, paperwork and workers reviewing the gap safely.
A late load can have several explanations. The warehouse may not have released it. The assigned vehicle may be unavailable. An appointment may have changed. Before buying an AI system, leadership should establish which explanation the available records actually support.
That is the useful starting point for this article: a proposed way to examine operating friction, rather than a claim about a deployed AI system. My recommendation is to begin with one repeated decision and follow the evidence through the departments involved.
Follow the handoff
Consider a hypothetical load that dispatch marks ready while maintenance still shows the assigned vehicle on hold. A useful review would display both records, their timestamps and their owners. It would ask for reconciliation before recommending a dispatch action. A simple rule may identify that contradiction perfectly well. An AI assistant might help summarize the surrounding messages, provided a reviewer checks the summary.
The distinction matters because generative systems can confidently produce false content, a risk identified in NIST’s generative-AI profile. [1] Fluent explanations should therefore remain separate from the underlying operating record.
For this pilot, I would require an evidence trail: the original status, the most recent update, the person authorized to correct it and the action taken. A system that cannot distinguish an old status from a current one should stop and request confirmation.
What my operating record supports
My Career Vault records fleet uptime improving from 61.4% to 95.9% and outside-carrier spending falling from approximately USD 2 million to USD 300,000 per month at PetroChoice / MOOVE. These are author-controlled career figures, checked against that record rather than independently audited employer data. They are operating-management results; I do not attribute them to AI. [2]
Their relevance is limited but practical. They identify the kinds of measures I would want an operating review to make understandable. They do not establish that the proposed assistant could reproduce the results, or establish which intervention caused each historical change.
Test the smallest useful intervention
I would first review a sample of exceptions manually. Record where evidence conflicts, how long reconciliation takes and whether the same problem returns. Then compare the existing process, a simple rules-based queue and an AI-assisted version using the same case definitions.
Count reviewer corrections and missed cases as well as time saved. Include integration, checking and support work in the cost. If better data ownership solves the problem, keep that improvement even if the AI component adds nothing.
NIST’s voluntary AI framework supports context-specific evaluation and accountability. [3] It does not supply a business case for this pilot. That case must come from the operation being tested.
The first useful result may be an uncomfortable finding: nobody owns the disputed status. Software can help expose that gap. Leadership still has to assign the authority and follow through.
SOURCES & EVIDENCE
[1] Chloe Autio et al. / NIST (2024-07-26). Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf. Locator: NIST AI 600-1, section 2.2 and action MS-2.5-003. Supports: The risk of confidently false generated content. Limitations: Identifies risks and suggested controls; it does not validate these proposed workflows. Checked: 2026-10-17.
[2] H. Max Ottesen (2026-10-15). Executive Career Vault and Resume Rules. CAREER_VAULT_SOURCE.pdf, Version 3 Final, sections 5–7, pp. 2–4. Locator: CAREER_VAULT_SOURCE.pdf, Version 3 Final, sections 5–7, pp. 2–4. Supports: The two PetroChoice / MOOVE operating figures, with attribution and audit limits. Limitations: Verified against the governing Vault only. Employer records, underlying calculation definitions and independent audit evidence were not supplied. Checked: 2026-10-17.
[3] Elham Tabassi / NIST (2023-01-26). Artificial Intelligence Risk Management Framework (AI RMF 1.0). https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf. Locator: NIST AI 100-1, Core, sections 5.1–5.4. Supports: Context-specific evaluation and accountable governance. Limitations: A governance framework, not evidence that a proposed operating tool improves results. Checked: 2026-10-17.
HOW THE CONCLUSION WAS REACHED
The career record identifies relevant management measures; the NIST material identifies evaluation and reliability concerns. From those premises, a small comparison against the current process is a reasonable test design. Neither source proves that AI improves handoffs.
LIMITATIONS
No AI trial, cost study, employer source data or causal attribution was supplied. The warehouse example is hypothetical. The title describes a conditional possibility, not a finding that every broken operation can be diagnosed by AI.
© 2026 H. Max Ottesen. All rights reserved.
← All articles & insights