# First AI workflow card Fill in one card for the first AI workflow you plan to run. Fill in what you know; every blank line gets an owner and a date. Any blank line means: not ready to run. | Question | Answer | Owner / by when | | --- | --- | --- | | Which repeating work goes first: documents, quoting, dispatch support or track and trace? | | | | Volume: how much of it lands each week, and how many people do it? | | | | What does an error in that work cost today, and is the worst case a correction? | | | | Boundary test: does the output move a load, change a rate a shipper sees, watch a driver or cross a border? (yes / no) | ☐ Yes ☐ No | | | If yes, whose bar applies (load, rate, driver, border), and which named person owns every output before it ships? | | | | Trained like a new hire and checked like a control, corrected hourly at first, then daily, then weekly? (yes / no) | ☐ Yes ☐ No | | | What already runs that nobody calls AI (TMS solvers, ETA models, yard systems)? | | | | What does that count change in the budget line? | | | | Which vendor claims have run on our freight: our lanes, our quote history, our documents? | | | | Price quoted, and what it buys? | | | | The number it moves: what moves, by how much, by when? | | | | Baseline for that number measured before the pilot starts? (yes / no) | ☐ Yes ☐ No | | | Who checks the number, and how often? | | | | Decision and date | ☐ RUN ☐ WAIT FOR THE BAR ☐ RE-TEST ON ______ Date ______ Signed ______ | | ## Example (fictional): status-email drafts at a 140-truck regional carrier Fictional example. | Question | Answer | Owner / by when | | --- | --- | --- | | Which repeating work goes first: documents, quoting, dispatch support or track and trace? | Drafting status replies to shipper "where's my load" emails; people review and send | n/a | | Volume: how much of it lands each week, and how many people do it? | About 450 status emails a week; three customer service reps handle them | n/a | | What does an error in that work cost today, and is the worst case a correction? | A wrong status is fixed in the next reply; no load moves and no rate changes | n/a | | Boundary test: does the output move a load, change a rate a shipper sees, watch a driver or cross a border? (yes / no) | ☐ Yes ☒ No; the reply reports where a load is, and a person sends it | n/a | | If yes, whose bar applies (load, rate, driver, border), and which named person owns every output before it ships? | Not needed at No; the customer service lead still signs every reply before it goes | n/a | | Trained like a new hire and checked like a control, corrected hourly at first, then daily, then weekly? (yes / no) | ☒ Yes; two leads correct drafts hourly for the first two weeks, then daily | n/a | | What already runs that nobody calls AI (TMS solvers, ETA models, yard systems)? | The TMS ETA model and the automated end-of-day status report nobody calls AI | n/a | | What does that count change in the budget line? | Feed the ETA model's output to the drafts instead of buying a visibility tool; about $40,000 less in the ask | Ops analyst, by 4 Dec | | Which vendor claims have run on our freight: our lanes, our quote history, our documents? | **Fails the bar**: the demo ran on the vendor's own sample emails, not our inbox | Vendor manager, by 11 Dec | | Price quoted, and what it buys? | About $60,000 a year for the email module, plus the TMS connection | n/a | | The number it moves: what moves, by how much, by when? | Reply time from about 4 hours to under 1 hour, and rep time on status emails from about 30 hours a week to 10, by end of Q1 | Customer service lead | | Baseline for that number measured before the pilot starts? (yes / no) | **Blank: not ready** | Ops analyst, by 4 Dec | | Who checks the number, and how often? | Customer service lead samples 25 sent replies a week and logs wrong statuses | n/a | | Decision and date | ☐ RUN ☒ WAIT FOR THE BAR ☐ RE-TEST ON ______ · Date 20 Nov · "No baseline yet, and the vendor has not run on our emails." | Ops director | From the course AI in transport and logistics, Dr. Tony Hoang · v1 · Oct 2026