Your reps re-key tenders all day. We draft the load. Your rep approves it.
Custom AI agents that work inside the inbox and TMS you already use. No migration, no new software for your team to learn. A person approves every action before it goes anywhere.
| Pickup | Fresno, CA · 11/03 · FCFS 06:00–14:00 |
|---|---|
| Stop 2 | Reno, NV · 11/04 |
| Delivery | Salt Lake City, UT · 11/05 |
| Equipment | 53' Reefer · 34°F |
| Weight | 41,200 lb |
| Shipper ref | PO 88123 (found on page 2) |
| Delivery appt # | Missing: ask the shipper? |
Most of your back office is typing. None of it should leave without a person.
Tenders by email
A shipper sends a PDF with two stops, the PO buried on page 2 and “FCFS”. Someone types it into the TMS. Then does it sixty more times before lunch.
Carrier invoice audit
Rate con, invoice and POD live in three places. $150 of detention nobody agreed to slips through, or the invoice sits for a week.
Quote requests
RFQs pile up in a shared inbox while the lane goes cold. Your best rep spends the morning copying origin, destination and weight into a spreadsheet.
The agent does the typing. Your rep keeps the decision.
Connect the inbox
We read the mailbox where tenders and invoices already land, in Gmail or Outlook. Nothing changes for your shippers or carriers.
The agent drafts and flags
It pulls the fields, builds the load or checks the invoice, and flags what’s missing or doesn’t match: no appointment number, a charge that isn’t on the rate con.
Your rep approves
One click sends it to the TMS. Corrections teach the agent. Every action is logged with who approved it.
- Aljex
- McLeod
- Tai
- Turvo
- Revenova
- Rose Rocket
- Gmail
- Outlook
The same pattern, already running at a global payments company.
Before starting Dowel, Nicolás Gut built the agents behind a global payments company’s internal service desk: requests come in, an agent drafts the resolution, a person approves it before anything goes live, everything is logged. Different industry, same shape as a tender becoming a load.
- 15 → ~4 min
- handling time per request
- ~50%
- of requests resolved by the agent, with a person approving
- 0
- changes the agent puts into production on its own
A 5-day Diagnostic before anyone builds anything.
We start with the work, not with the AI. Pick one process (tenders, invoice audit or quotes) and we sit with the people who do it.
- We time 20–30 real cases with your team.
- You get the minutes per item today and what an agent would take off your reps, in hours and dollars.
- You get a fixed-price proposal for a pilot, with the success criteria written down.
One process · a fixed fee, credited in full if you go ahead with the pilot · you keep the numbers either way
Book a 20-min callA small studio, led by Nicolás Gut
Dowel builds AI agents for operations teams. We don’t sell a platform: we build the agent your process needs, on top of the tools you already pay for, and we stay accountable for it after it ships.
Nicolás is a software engineer who has built agents for operations at a global payments company. We work US business hours, and you talk to the people who build it. LinkedIn
Do we have to change TMS?
No. The agent works through your inbox and your TMS’s import or API. Your reps keep the screens they know.
What happens to our data?
It stays in your accounts. We don’t train models on it, and you can revoke access at any time.
What if the agent gets it wrong?
Nothing reaches a shipper, carrier or your TMS until a person approves it. Mistakes become corrections, not incidents.
Bring one process that eats your reps’ day.
Twenty minutes, no deck. If an agent can’t save you real hours, we’ll tell you on the call.
or write to [email protected]