Custom agents, built around how you work
No generic "skills" to configure. We map your real workflow, then build an agent team with your rules, your exceptions and your customers' quirks.
Service, not a seat licenceWe build custom AI agents for
freight forwarders
— wired into the legacy systems you already run, trained on your own data, and live in weeks. Your people approve. Agents do the rest.
Decision-intelligence and planning suites are built for global enterprises with clean data lakes and multi-quarter programs. Most forwarders, carriers and 3PLs run on email, PDFs, EDI and a TMS that's older than their newest hire. That's exactly where we start.
No generic "skills" to configure. We map your real workflow, then build an agent team with your rules, your exceptions and your customers' quirks.
Service, not a seat licenceEDI, SFTP, AS/400, ODBC, Excel drops, shared inboxes and green-screen TMS. If a human can read it, an agent can work with it — no rip-and-replace.
Start without a data lakeOpen-weight models deployed in your VPC or on-prem, adapted to your codes, lanes and document history. Your data never trains someone else's model.
Open-weight · on your hardwareAgents don't stop at recommendations — they post to the TMS, send the update, book the slot. Low-confidence fields, unknown layouts and money decisions stop for a human, with the evidence attached. Every action is logged and reversible.
Execution with guardrailsOne workflow, scoped in days, in production-like test in about a week. Expand once the numbers prove it.
5-day reference pathThis is the agent team we built for a freight forwarder, now running in production. It quotes, reads Outlook and Teams in real time, checks documents, applies each customer's SOP, works inside their legacy freight ERP, keeps one trusted shipment record, chases carriers and invoices on proof of delivery. Click an agent to see what it does and where a person stays in control.
Your team asks in plain language: search, draft, book, status, complete. The Orchestrator works out the intent, brings in the right specialist agents, runs their checks in parallel and returns one clear answer, with the next action and who owns it.
Your team asks in plain language: search, draft, book, status, complete. The Orchestrator works out the intent, brings in the right specialist agents, runs their checks in parallel and returns one clear answer, with the next action and who owns it.
Explore the Orchestrator Agent →
Pre-booking salesThe Quotation Agent turns quote requests from email and Teams into priced, customer-ready quotes. It looks up contract and carrier rates, applies each customer's pricing rules and margins, drafts the quote for approval, chases open quotes and hands accepted ones straight to booking.
The Freight Agent turns emails, Teams messages and attachments into evidence on the right shipment. It checks documents against the stage you're at, triages exceptions and proposes the next action — with severity, owner and due time.
The Document Agent works out which documents a shipment needs for its mode, route, customer and stage: commercial invoice, packing list, bill of lading, air waybill, customs declaration, certificate of origin, insurance. It checks what has arrived, cross-checks the details and chases what's missing before the cut-off.
Each customer has its own way of shipping. The SOP Agent reads your customer SOP workbooks from Teams and SharePoint, tracks every version, picks the right variant for each shipment and checks bookings against it before anything moves.
Most forwarders run a freight ERP that won't be replaced this year. The Legacy ERP Agent works with it directly through its own service interface (SOAP/WCF in our first deployment). It previews every change, waits for exact approval, executes, and reads the result back.
Explore the Legacy ERP Agent →
Digital twin & control towerThe TMS Agent keeps the single source of truth for every shipment: references, timeline, exceptions, actions, owners and documents. It tracks operational KPIs, flags delayed departures and arrivals with their impact on manufacturing and warehouse distribution, and gives executives shipped and arrived volumes against forecasted sales.
The Partner Agent manages connections to carriers, airlines, terminals and other partners. It resolves the right tracking identifier, reads the latest status from official sources, registers tracking and normalizes partner events into your timeline.
The Invoicing Agent closes the loop after delivery. It collects proof of delivery, reconciles charges against the quote, booking and carrier invoices, drafts the customer invoice in your billing system, flags disputes and follows up until payment arrives.
One orchestrator coordinating eight specialists, from quote to cash.
Messaging and acknowledgement across Outlook and Teams, so every action has an owner.
search, draft, book, status, complete. That's all your team needs to learn.
Records created in the freight ERP without exact human approval.
Supply chain runs on systems nobody wants to migrate. We connect agents to what you already have — through APIs where they exist, and through files, feeds and screens where they don't.
PDFs, scans, emails and spreadsheets turned into structured, validated records.
EDI, SFTP drops and flat files read and written on your existing schedules.
Read-only by default over ODBC, SQL, REST or SOAP; writes only through approved paths.
Where there's no interface, agents operate the screen a person would — logged and supervised.
System names are trademarks of their owners and are listed as examples of integration targets, not partnerships.
Shipping data is commercially sensitive. We can deploy open-weight language models inside your environment and adapt them to your documents, codes and history — so the intelligence you build stays an asset you own.
We sit with the team, collect real samples, and pick the slice with the most retyping and the clearest rules.
We build the agent team, connect your systems and set the guardrails your people want.
Production-like test on your real documents. Measured against baseline effort, cycle time and rework.
We run, monitor and improve the agents — then add the next workflow your team chooses.
Reference engagements with the details your operations and IT leads will ask about.
Order PDFs extracted, validated, enriched with port codes and delivered for TMS import, with exceptions routed to people.
Read the use case → Freight forwarding · agent teamA nine-agent team for a freight forwarder, from quote to cash: inbox, documents, customer SOPs, legacy freight ERP, TMS, carriers and invoicing. In production in 8 weeks.
Meet the agent team → Commerce workflowAn undocumented legacy system mapped, extended and handed over with rollback and admin manuals.
Read the use case →No. We're a service. We build custom agents around your workflows and systems, then operate and improve them with you. You own the workflows, rules, mappings and data.
Those are enterprise platforms for decision intelligence and planning. We build execution agents for the day-to-day operational work — documents, bookings, updates, exceptions — inside the systems mid-market logistics teams already run. See comparisons →
Yes. Agents can work through EDI, SFTP, file imports, database reads and, where needed, supervised screen automation.
It doesn't have to. We can deploy open-weight models on your infrastructure, or route only non-sensitive tasks to hosted models you approve.
Confidence thresholds, business rules and approval gates. Uncertain items stop for a person with the evidence attached, and every action is logged.
Yes. Our TMS Agent flags delayed departures and arrivals and shows the impact on manufacturing schedules and warehouse distribution. It also flags capacity gaps, inventory mismatches and missing documents. Exception checks across logistics and warehouse inventory are acknowledged by your team, so nothing closes silently. Executives see what shipped and arrived against forecasted sales.
It's designed to let the same team handle more volume and spend time on customers and exceptions instead of retyping.
In one call we'll tell you whether agents fit, what the first slice should prove, and which systems we need to touch.