Skip to content
Kyle Rivelli.

Work

Email bots · Shadow: drafts, a human sends · since July 2026

The carrier negotiation bot.

Haggles with carriers by email. Every dollar comes from a deterministic ladder; the AI only reads.

Shadow: drafts, a human sendsClaudeMicrosoft GraphFMCSA registryTMS APIstate machine
inboxAI readscode decideshuman sends

The problem

Carriers email about posted loads all day and haggle. Someone answered every one by hand, and the after-hours ones went unanswered entirely. Industry job boards list $45,000 to $60,000 a year roles for exactly this work.

How it works

  • Every dollar figure a carrier ever sees originates in a pure pricing module with a fixed ladder and a hard floor. The AI can only phrase numbers this module returns; it cannot invent one.
  • Claude reads each inbound email and produces structured intent (an ask, a counter, an acceptance, something off-script) for about $0.003 per email. It never writes the reply and never chooses the route.
  • A pure state machine runs the protocol. Agreement is the only outcome that reaches a human. A negotiation that fizzles dies politely. Anything off-script parks for review.
  • Carrier vetting is a deterministic chain: pattern extraction of MC and DOT numbers, the federal carrier registry, then the TMS. The ambiguous middle parks for a human.
  • A margin cap added in October: the bot never offers more than 96 percent of what a load sells for. A load that cannot be covered inside that line is declined politely and flagged to the desk as a loss load, never quietly dropped.
  • Runs as an always-on worker on a desk PC, work hours only, with push notifications for the cases that need a person and automatic hand-off to the brokers who finalize.

// the AI never touches a price; every dollar comes from a deterministic ladder, capped at 96 percent of sell

The result

3,808 carrier threads handled in shadow mode, where the bot drafts and a human sends: 243 reached agreement, 1,900 were declined cleanly, 546 were parked for a person. Roughly forty behavior refinements came from watching it against live traffic before it was trusted to draft on its own.

In practice

Synthetic example: a thread ending at the human gate

Schedule of the unattended systems, Central time.

Syncing the Central clock.

Your version of this exists.

Different industry, same shape: a repetitive workflow, a fragile spreadsheet, an inbox someone babysits. Tell me what the manual work looks like and I can usually name the first build in one conversation.