Spartan trials AI-built hourly customer route reports
What the company did
Spartan Carrier Group built an Alvys Foundry AI agent to prepare hourly route-status reports in customer-requested formats. Dispatchers had assembled each report by hand in Excel. Working with Alvys staff, the carrier’s operations lead configured the agent to generate the reports inside its existing transportation management system.
How it went
FreightWaves’ October 27 report draws on a September interview with operations lead Kendrick Dawson. He said the agent was running within a week and a half and reports arrived in his inbox on the hour, matching manual versions almost completely. The former task averaged seven to ten minutes per report; he estimated it could free 17% of a dispatcher’s day. Customer delivery was still planned at the interview. No comparison sample size or independently verified savings were supplied.
Worth knowing
- Alvys staff built the first agent with Dawson. Its template needed a customer name and report-format choice, but the article does not publish the exact shipment fields, model, prompts or data-freshness controls.
- Practical takeaway: Compare each report with an approved manual version before customer distribution. Check shipment references, latest milestones, timestamps and time zones, and unresolved exceptions; a report arriving on the hour can still contain stale shipment information.
Sources & provenance
FreightWaves · Published 2026-10-27 · Reviewed 2026-10-28
“Its first AI agent builds the hourly customer reports that dispatchers had been making by hand.”
Supporting evidence and review
Human review required
- Compare generated reports with the approved manual version before enabling customer distribution.
- Check shipment references, latest milestones, update timestamps and time zones, and unresolved exceptions; do not treat an on-time report as proof that its data is current.
Evidence & limitations
- Original trade interview with the carrier’s operations lead; AI & Freight has not independently tested the agent.
- The October 27 article reports a September interview. Internal generation and comparison are described; customer delivery was planned for the following Monday, not independently confirmed in the report.
- The estimated 17% of a dispatcher’s day is a manager’s estimate, not an audited labor-saving measurement.
- The reported near-complete match to manual reports has no disclosed sample size, scoring method or exception rate.
- Exact model, prompts, report fields and data-freshness controls are not published. A separate driver-payroll workflow was untested and its projected savings are excluded.