AI for logistics teams, from route optimisation to shipment anomaly detection.

Generally no. Most logistics AI - route optimisation, demand forecasting, document automation, anomaly detection on shipment data - sits outside the EU AI Act's high-risk classification. The exceptions are specific and we flag them at scoping: AI used as a safety component of a transport system (Annex III §2), AI used in driver monitoring, scoring or recruitment (Annex III §4), and AI used by public authorities for customs or border control (Annex III §7). Where one of those triggers fires, we build to the high-risk obligations from week one. Where they do not, we build proportionate controls instead of pretending the higher tier applies.

Driver and warehouse-worker data is personal data under GDPR, including identifiers tied to a vehicle a specific driver operates. We document the lawful basis for each processing category, minimise the data the AI sees to what the use case actually requires, sign DPAs with zero-retention and no-training clauses for inference traffic by default, and produce a data-protection memo before any system goes live. For Article 22 territory - driver scoring, automated hiring, contract triage - we design the human-in-the-loop step into the workflow, never as an afterthought.

Yes. The delivery layer is built around your stack, not the other way around. We ship integrations with the major TMS platforms (SAP TM, Oracle Transportation Management, Manhattan Active TM, Blue Yonder TMS, MercuryGate, Alpega), WMS platforms (SAP EWM, Manhattan Active WM, Blue Yonder WMS, Körber), forwarder operating systems (CargoWise, Descartes, Riege Scope), and visibility platforms (project44, FourKites, Shippeo, Sixfold). For systems without a modern API we build a queue-based bridge with idempotent writes and a manual reconciliation interface. Identifier handling follows GS1 keys (GTIN, SSCC, GLN) and event semantics (EPCIS) so the AI interoperates cleanly across partners.

Yes, and this is the difference between a model that sticks in production and one planners override into uselessness within six months. Every reroute the system proposes is delivered with the constraints that were active (capacity, time windows, driver hours, hazmat rules), the input features that drove the choice, and the alternative routes considered with their cost. A planner who needs to defend the reroute to a customer or an ops controller can do it in under a minute. Where overrides happen, the override reason feeds back into the eval set so the next model version learns from the correction.

Not on this engagement. Customer-communication automation under this scope means digital channels: email, EDI 214 status messages, customer portal replies, partner status APIs, and chat surfaces. Voice-AI is a separate product domain with its own latency, telephony and recording-compliance requirements. If you want a voice channel, we will be honest about whether it is the right fit and refer you to a specialised team rather than ship it inside a logistics-AI engagement.

A first engagement targets one workflow with a measurable baseline, runs 4 to 12 weeks to production, and lands as a single signed-off system inside one TMS, WMS or operational surface. Common first scopes are: route optimisation across one lane group; demand forecasting across one product family; document automation for one document type (BOL or customs declaration) on one trade lane; anomaly detection on one carrier or one mode. Submit a project with the workflow you have in mind and the rough volume, and we scope and price the discovery phase before any code is written.

By default, processing and storage runs in EU regions on infrastructure under EU jurisdiction. We support regional pinning when a regulator or customer contract requires it (Germany-only, France-only, Lithuania-only, US-only). Original shipment data lands in immutable EU object storage with hashes recorded in the audit log. We do not train any model on your shipment, customer or driver data. If your contract requires US-resident processing for a US-only lane, we expose that as an explicit configuration toggle, never a default.

Pricing is set after the discovery sprint, against your specific workflow, lane mix and integration surface. We do not publish a flat rate because the scope variation across logistics AI is wide: a single-lane document-automation system on a uniform corpus is a different build from a network-wide demand-forecasting and anomaly-detection stack across a multi-region 3PL. Submit a project with the workflow and rough volume, and we come back with a discovery proposal within one business day.