/ AI for logistics

The office should not be a day behind the road.

A delivery finishes, the driver moves on, and the job stays open in the office while somebody locates the proof and reconciles the exception. That gap sits upstream of customer updates, invoice release and cash, which is why closing it is worth more than any individual efficiency.

A regional same day and next day freight operator

Under three million pounds in turnover, moving 35 to 55 consignments daily.

The problem

Proof of delivery arrived as photographs, scans and messages with inconsistent references. Exceptions were reconstructed by hand, so the office was always behind the road and invoices waited on evidence somebody had to go and find.

After

Evidence is matched to the consignment as it arrives, exceptions are classified and the open list is visible in one place. A person still decides liability, agrees a credit and releases an invoice.

Around 16 hours a weekRead the full engagement

A delivered Orchyn build. Clients are anonymised at their request, so the sector, scale and measured outcome are real and the name is withheld.

Where the time usually goes

Proof of delivery in five formats

Photographs, scans, signatures and messages arrive with references that do not match the consignment record, so matching is manual.

Exceptions reconstructed after the fact

By the time somebody assembles what happened, the customer has already asked twice and the invoice is still unreleased.

Carrier invoices that do not match the rate

Surcharges inflate, charges duplicate, chargeable weights and free time get misread. Each error is small enough to wave through, which is why they are absorbed month after month.

Cash waiting on paperwork

Invoice release depends on evidence being found. The delay is administrative, not commercial, but it costs the same.

Where the system stops

It does not decide who is liable for a failed delivery, agree a credit or a claim, release an invoice or make a promise to a customer. It matches evidence and surfaces what is still open. Fixing reference conventions across the operation did more work than the model did, which is typical.

This is the house position across every Orchyn build: AI assisted, never AI executed. A model is used where judgement over messy input is needed. Every figure is computed by deterministic code against a named source, and a person holds authority over anything with commercial or legal consequence.

Common questions

01What is freight invoice audit?+

Checking every carrier invoice against the rates you agreed and against published market rates, then reporting the exact amount you were overcharged. FreightGuard, one of our own builds, does this across road, pallet, air and ocean. The model identifies the kind of problem, such as an unfamiliar invoice layout or a charge that does not belong, and deterministic code computes every figure against a named source. That division is what makes a finding defensible in front of a carrier: the arithmetic can be reproduced line by line without trusting the model at all.

02Can you show us what it recovered for a client?+

No, and we would rather explain why than give you a number. Client figures are confidential, so the figures shown in our public material are representative rather than real recoveries. We can walk you through the method in detail and show you the evidence pack format, which is usually more convincing than a headline figure you cannot verify.

03Would it talk to our customers?+

Not unless you decide it should, and in the delivered build it does not. It prepares the picture so your team can update a customer faster. Anything that commits you to a customer, such as agreeing a credit or accepting liability, stays with a person.

04Does it need to replace our transport management system?+

No. Where existing systems remain valuable we design around them and improve how information moves between them. The freight build sat alongside the operator's existing systems rather than replacing them.