Every Cortex report, turned into the right text to the right driver.
MJS Driver Messaging reads the daily Amazon reports, writes one clear coaching message per driver in their own language, and puts it in front of a manager before anything is sent. Hours, breaks, safety events and the daily GM and compliance reports come from the same data.
Managers export the same reports they already pull from Cortex and ADP. The tool does the reading, the writing, and the bookkeeping.
Missing packages, returns without a reason code, customer defects, pace and safety events are merged into a single numbered message. A clean day gets a thank-you instead.
Every message waits in a review queue. Edit it, approve it, send it. Replies come back into the same record, marked unread until someone reads them.
Paid hours against a 10-hour budget, the 30-minute lunch window, split shifts, rest between shifts and the weekly limit, checked from the closing report and the ADP timecard.
Netradyne camera events are tied to the driver who had the van that day. Yard moves before departure are set aside, so nobody is coached for an event that was not theirs.
The Daily GM Operations Report and the Compliance Manager report fill themselves in from the data, with the evidence written next to each check. Print them or keep them on screen.
Build tomorrow's schedule the night before and watch it reconcile against what actually happened. See each driver's projected scorecard standing for the week, and ask the data any question in plain English.
The same routine every day, from the night before to the end-of-week review. No connection to Amazon systems; managers export the files by hand, which keeps the station inside Amazon's rules.
The dispatcher sets tomorrow's waves, drivers and vans, marks who confirmed, and notes call-outs, stay-homes and on-call. Van conflicts and drivers near the weekly hours limit are flagged as they type. Routes go in the next morning when Amazon assigns them.
Routes, Packages, RTS Quality, Customer Feedback, the closing report, the ADP timecard and the safety export. The tool works out which station and which day each file belongs to, and the checklist shows what is still missing.
One line per driver with the issue tags, the message in English and in the driver's language, and the facts behind it. Approve, edit, or hold. A driver never gets a "great day" text at noon and a "missing package" text at five.
Messages go out from the station's own phone line. Replies, delivery receipts and the coaching history build up per driver, with new replies flagged until a manager reads them.
The Daily GM Operations Report and the Compliance Manager report fill themselves in from the same data: hours, lunches, no-shows, fleet, safety, disputes. The GM adds what no export knows and prints it.
Each driver gets a short plain-language appraisal of their week or month, written by AI from their own numbers, checked by a manager, and sent the same way as any other text.
The daily coaching texts are fixed templates, so they say the same thing every day. AI is used where judgement and writing are needed, and every one of those outputs stops for a person before it goes anywhere.
Claude writes each driver's review from their own pace, missing packages, returns, customer feedback and how their standing moved, in the driver's language. It lands in the approve queue like any other text and never sends on its own.
When Amazon raises a finding, Claude gathers the scans, route, clock-in, feedback and safety events behind it and writes the dispute with the tracking IDs and timestamps. It rates the case strong, fair or weak, and says plainly when the data does not support it.
"Who went over ten hours three times this week?" Claude turns the question into a read-only query, runs it, and explains the rows. A floating helper on every page. Managers only ever see their own station, no matter how the question is phrased.
Guardrails: nothing AI-written is sent or submitted without a manager approving it, AI queries can only read, and the model is given the station's facts, never asked to invent them.
Each text is sent in English first, then in the language set for that driver. Templates are written by people, not translated on the fly, and weekly reviews are written by an AI assistant and approved by a manager before they go out.
The live sample runs every page on a sample station fixed on one day, no login needed. Managers at MJS sign in to the dashboard.
Open the live sample Manager sign in