The intelligence layer for last-mile delivery

What one driver learns about a route, the next one inherits.

Drivers know things no map does — which bell works, where the van fits, when the gate is locked. ParcelVox captures that knowledge by voice, AI turns it into a structured, location-tagged tip, a dispatcher gives the final approval, and it plays back to the next driver at exactly the right stop.

M. Kaur · Route DE-1184 · Stop 24/61
“Front bell is broken. Ring at the side door in the yard — code 4711.”
Voice note · hands-free
AI structuring · Berlin Nord Hub dispatch
Structured as access · stop 24 — side entrance, gate code, matched to the stop record. Ready for a one-tap check.
Dispatcher approved · human in the loop
Next driver · Same route, next run
Arriving at stop 24: “Heads up — the front bell is broken. Use the side door in the yard, code 4711.”
Inherited tip · plays at the stop
Illustrative demo — how one tip travels
The problem

The best route data isn’t in your system.
It’s in your drivers’ heads.

Parcel logistics is highly automated — until the driver leaves the depot. From there, delivery depends on what the driver knows: where to park, which entrance works, how to reach the right door. That knowledge is captured nowhere, so the carrier pays for it three times over.

Street at dusk with a parked delivery van and labeled friction points: broken bell, unknown gate code, blocked dock, no-stopping zone BROKEN BELL GATE CODE? DOCK BLOCKED NO STOPPING 8–17

The friction drivers handle every day — invisible to the planning system

Ramp-up and churn

A new driver needs four to six weeks to reach full speed, and many don’t last that long. The knowledge leaves with them; the next hire starts from zero.

The same friction, relearned

Every driver rediscovers the entrances, parking spots and doors the last one already found — repeated across thousands of stops.

No insight into why

Dispatch sees that a stop ran late or failed, but never why. So tomorrow’s plan repeats today’s mistakes and the ETAs stay wrong.

When a driver hands in their keys, years of route knowledge walk out the door.

How it works

One simple loop. Three steps.

Step 1 · Capture

The driver says it

Drivers talk to a voice assistant — hands-free, while they work. No typing, no forms. It can also notice when something looks off, like an unusually long stop, and simply ask what happened.

Step 2 · Validate

AI structures it, a human approves

AI turns the voice note into a structured, location-tagged tip and matches it to the stop record. A dispatcher gives the final approval in one tap — human in the loop, minimal manual work.

Step 3 · Inherit

The next driver hears it

On the next run, the tip plays at the right stop, at the right moment. New drivers start with the knowledge of everyone who drove the route before them — and each shift sharpens what the assistant asks next.

STRUCTURED FRICTION IN  →  BETTER ROUTES OUT

Approved tips do a second job: they feed back into route planning as structured, geo-referenced records — helping improve ETAs, stop order, and first-attempt success over time.

For drivers

Built with drivers, not against them.

ParcelVox is about what drivers know — never about how fast they drive.

A driver in a hi-vis vest beside a delivery van speaks a voice note; AI structures the tip, a dispatcher approves it, and it is passed on to the next driver AI STRUCTURES IT DISPATCHER APPROVES NEXT DRIVER HEARS IT

Speak it hands-free → AI structures it → a dispatcher approves → the next driver hears it

No speed scores

Nothing here counts stops per hour or ranks drivers against each other. This is not a performance tool.

About places, not people

Tips describe stops and buildings — a broken bell, a safe parking spot — not the driver.

For carriers

What your operation gets

Faster ramp-up

New and temporary drivers inherit the route from day one, instead of spending four to six weeks learning it the hard way.

Known problems stop repeating

A broken bell or blocked access gets captured once — so the next driver doesn’t hit the same wall, and neither does the one after that.

A knowledge base you can trust

AI structures every tip and matches it to the stop record; a dispatcher gives the final approval. Trustworthy data, almost no manual work.

Planning that learns

Dispatch finally sees why a stop ran late or failed, not just that it did. Structured records feed route plans and ETA accuracy.

Ships inside your own app

Delivered as an SDK in the driver app your drivers already use — no second app to roll out. Sold per depot.

A data asset that compounds

Approved tips accumulate into a corpus a competitor can’t buy. It only grows by running real deliveries.

A sortation hub sends routes out to three doors; approved tips and friction data flow back into route planning SORTATION HUB TIPS + FRICTION DATA ROUTE PLANNING

From the hub to the door — and the knowledge flows back into planning

Team

Who is building it

Two founders, backed by Beam.

Vincenzo De Salvo · Commercial

Supply-chain and logistics background from Sony and Tetra Pak in Sweden. Later founded his own AI company before joining Beam to build ParcelVox.

Johannes Berding · Technology

Founded a travel-tech startup built on a novel routing algorithm. Brings GIS algorithms and device sensor signal interpretation — GPS, motion patterns — to the ParcelVox platform.

Contact

See it on your routes.

We’d rather show you than tell you. Tell us a little about your operation and we’ll set up a walkthrough — how tips get captured, checked and passed on, and what integration in your own driver app would look like.

This form isn’t connected to a backend yet — wire it up before publishing.