AI Agents & Workflows for Logistics and Freight Forwarding
Bills of lading, commercial invoices and packing lists arrive as scans and become manual data entry. We build the workflows that read them, check them against each other, and escalate to your ops team only when something genuinely does not add up.
The work we take off the desk
Named processes, not categories. If none of these are your bottleneck, the assessment will say so.
Documents re-keyed by hand
Bills of lading, commercial invoices, packing lists and certificates of origin arrive as email attachments. Operations staff read them on one screen and retype them into the forwarding system on another, all day, for every shipment.
Mismatches found too late
The invoice total does not agree with the packing list. The container number on the arrival notice is not the one on the booking. These are caught downstream, after a declaration has been filed or a charge has been passed on.
Quotes sitting in an inbox
An RFQ waits while someone finds the right contract rate, checks which surcharges apply this month and whether the validity window has closed. The lane is standard; the delay is not.
Status chased manually
"Where is my container?" is answered by logging into three carrier portals. The information exists; assembling it is the job.
What we actually build
Every step below is a real step in the build. Where a person still decides, it is marked — those gates are designed in, not bolted on.
Shipment document intake
The highest-volume, most mechanical task in the office, and the one that pays back first. Deterministic path, enumerable exceptions — a workflow, not an agent.
- 1
Classify every attachment
Bill of lading, commercial invoice, packing list, certificate of origin, arrival notice, delivery order. Each classification carries a confidence score that later steps act on.
- 2
Extract the fields that matter, per document type
Shipper and consignee, container and seal numbers, HS codes, gross weight, volume, Incoterm, vessel and voyage, ETD and ETA, charge lines. Different document types get different extraction schemas rather than one generic pass.
- 3
Cross-check the set against itself
Invoice total against packing list. Container number against booking confirmation. Declared weight against tolerance. A single document that reads cleanly is not the same as a consistent shipment file.
Confidence gate
Human decidesClean, internally consistent sets continue automatically. Anything below the threshold, or any mismatch, routes to a person with the specific field highlighted and both source images side by side — not the whole PDF to re-read.
- 5
Post to the forwarding system with an audit trail
Written via API where one exists, or a supervised UI automation where it does not. Every field records where it came from, which document, which page, and whether a human touched it.
Rate quotation assistant
Requests arrive in free text, missing fields, in several languages. That is exception handling, which is where an agent earns its cost over a fixed workflow.
- 1
Read the request and pull out what is quotable
Lane, ready date, volume and weight, Incoterm, cargo type, hazardous or temperature requirements, service level.
- 2
Identify what is missing and draft the follow-up
In the language the enquiry arrived in. The agent decides what it needs rather than following a fixed field list, because what is missing differs every time.
- 3
Query rates, surcharges and validity
Contracted rate tables, current surcharge schedules, validity windows. Lanes with no live contract are flagged rather than guessed at.
- 4
Assemble the quote with a charge breakdown
Line-by-line, with the validity date and the assumptions it used stated explicitly.
An account manager approves before anything sends
Human decidesPricing leaves the building only with a human name on it. The agent drafts; it does not commit the company.
ETA & exception monitoring
Turns status chasing into proactive notification, which is the part customers actually notice.
- 1
Poll carrier and terminal data for active containers
Across whichever portals and feeds your carriers actually publish to.
- 2
Detect what has changed materially
ETA slippage past your threshold, rollovers, customs holds, and demurrage or detention windows that are about to open.
- 3
Rank by customer impact and cost exposure
A two-day slip on a routine lane and a hold on time-critical cargo are not the same alert.
Draft the customer notice; ops approves and sends
Human decidesNew ETA, cause where known, and the options available. Sent by a person, so the relationship stays with your team.
How you will know it worked
We are a young practice and we do not have a wall of client logos to point at. So instead of asking you to trust results you cannot check, here is exactly how the result gets measured on your data — and how you check it yourself.
The metric is agreed before anything is built
You and we write down what is being measured and what counts as good, in advance. If we cannot agree a measurable definition, that is a signal the process is not ready to automate — and we would rather find that out in week one.
The baseline comes from your records, not ours
The "before" figure is drawn from your own system logs for the period preceding go-live. No industry benchmark, no vendor-supplied comparison, no number we brought with us.
Measurement runs in production, on your data
Live operation over an agreed window, against the documents and cases you actually receive. Not a curated test set, and not a demo environment where the inputs were chosen by us.
You get the raw log, not a summary
Per-item results including every case the workflow got wrong and why it went wrong. You can recompute our headline number yourself, and we would rather you did.
The result is reported either way
Including when it falls short of the target we agreed. A supplier who only reports the wins is not measuring anything — they are selecting. You will see the misses in the same document as the hits.
It runs inside the systems you already have
Nobody logs into a new tool. Where an API exists we use it; where one does not, we use supervised UI automation against the same screens your staff use — and we tell you which is which before you commit.
How a build runs
Discovery
We sit with the people doing the work and map the process as it actually runs, including the exceptions nobody wrote down. We come back with a shortlist of candidates ranked by volume, error cost and how cleanly they can be automated.
Pilot
One workflow, built end to end and put into production against your live systems. We agree an accuracy target up front and measure against your data, not a benchmark set. If it misses, you see the number.
Deploy
Integration hardening, access control, audit logging and the human escalation paths. Your team is trained on the runbook and owns the operating procedure before we step back.
Operate & extend
Monitoring, drift review and tuning as your documents and edge cases change. Once one workflow is trusted, the next one costs far less than the first.
Start with one workflow
Not a platform rollout and not a strategy deck. One process, in production, measured on your own data — so the decision to do the next one is made on evidence.
What the pilot includes
- Discovery workshop: we map the process as it actually runs, not as the SOP describes it
- One workflow built end to end and integrated with your live systems
- An accuracy baseline measured on your own documents and data, published to you
- Handover documentation and an operating runbook your team owns
- 30 days of tuning after go-live
Built on our own stack
We are not assembling someone else's components. The engines underneath these workflows are the products we already build and run.
AI Knowledge Base
Document understanding is grounded in your own tariff files, SOPs and historical shipments, with a private deployment option so nothing leaves your environment.
Learn moreAIGC Engine
Customer-facing drafts written in the language the enquiry arrived in, in your house style.
Learn moreIT Consultancy & Integration
TMS, carrier portal and mailbox integration — including the supervised UI automation path for systems with no API.
Learn moreLogistics & Freight Forwarding: common questions
Our forwarding system has no API. Can you still integrate?
Yes. Where an API exists we use it. Where none exists we use supervised UI automation against the same screens your staff use, with the same permissions and a full action log. It is slower and needs more maintenance than an API, and we will tell you that during discovery rather than after.
What happens when the extraction gets something wrong?
It routes to a person before anything is written. Every extracted field carries a confidence score and a link back to the exact document and page it came from, so checking a flagged field takes seconds rather than re-reading the PDF. The workflow never silently posts data it is unsure about.
How accurate is the document extraction?
We will not quote you a headline number, because accuracy depends almost entirely on your scan quality and how varied your document set is. During the pilot we measure on your documents, agree a threshold with you, and publish the result — including where it falls short. A vendor quoting a single accuracy figure before seeing your paperwork is quoting someone else’s paperwork.
Where does our shipping data go?
Private deployment is available for the whole pipeline, and your data is never used to train models we deploy elsewhere. Document retention, access control and audit logging are configured to your requirements during deployment.
How soon is something actually running?
A pilot puts one workflow into production in 4-6 weeks. Document intake is usually the right first choice: highest volume, most mechanical, clearest payback, and it produces the labelled data that makes the next workflow cheaper.
Tell us the process that hurts most
The assessment is a working session, not a pitch. If the honest answer is that your bottleneck is not worth automating yet, that is the answer you will get.