Use case shipping · Liner operations · Hamburg port · English-language MLC

Bills of Lading + Filipino crewing - Decision Layer for Hamburg liner shipping in English

12,000 Bills of Lading daily under UCC + IMDG + EU MRV plus English-language MLC 2006 crew-change documentation for Filipino crews. Audit-ready before Hauptzollamt and flag-state inspections.

Workshop at Grindelberg All 7 use cases ←

Chapter 1 — The volume problem

A Hamburg liner shipping company processes 12,000 Bills of Lading daily.

A model German liner shipping company with 200 container vessels and on average 60 Bills of Lading per vessel per day moves around 12,000 customs-binding waybills daily. Each B/L has 5-80 line items. Each line item falls under at least three rule sets simultaneously: Union Customs Code (EU 952/2013), IMDG Code Amendment 42-24 (mandatory since January 2026), EU MRV emissions reporting (2015/757).

Manual processing per Bill of Lading: 8-15 minutes. Weekend cargo creates backlog through Wednesday. Three full-time case handlers occupied only with BL classification. Plus two for sanctions screening (OFAC, EU, BaFin) - each new list means manual review.

A Decision-Layer architecture splits these 12,000 daily transactions into three lanes: ca. 75 percent RULES (deterministic HS-code matching, weight tolerance, preference check, IMDG class matching), ca. 20 percent AI AUTONOMOUS (language variants of B/L from 60+ countries, confidence-based classification), ca. 5 percent HUMAN (dangerous-goods escalation, sanctions check, customs-audit cases).

At a Hauptzollamt inspection, what crosses the table is not the result, but the path. Audit trail per line item: rule version (e.g. imdg_42-24_2026-01-01), input hash, confidence score, escalation path, decision timestamp.

Chapter 2 — Decision-Record for a Bill of Lading

What a Bills of Lading decision-record looks like in practice.

Anonymised example of a Bills of Lading classification with Decision-Record structure. Each step with decider type, rule version, confidence score and audit trail.

BL-2026-05-17-09:42:17-DE-PHL

Container B/L Germany - Philippines · Received 17.05.2026 09:42 · 47 line items · Hamburg to Manila · MBL (no HBL chain)

Result Classified · 6 escalations
  1. 01 REGEL

    BL format + HBL/MBL detection

    BIMCO Standard Master Bill of Lading detected (no House BL chain in this transaction). Mandatory fields complete (Shipper, Consignee, Notify, Vessel, Voyage, Place of Receipt, Port of Loading, Port of Discharge, Place of Delivery). Rule bl_format_v2.1.

    ✓ Valid MBL
  2. 02 REGEL

    EORI + declarant validation (ATLAS/DAKOSY)

    Shipper EORI DE123456789012345 verified against EU EORI registry. Hamburg declarant linked to DAKOSY participant ID. ATLAS export message AES.E_USDA precondition satisfied. Rule eori_dakosy_v3.0.

    ✓ Validated
  3. 03 REGEL

    Sanctions screening (OFAC, EU, BaFin, UK OFSI)

    Shipper + Consignee + Notify screened against 4 lists with fuzzy-match threshold 0.92. 1 weak match on Consignee (similarity 0.93, Philippines-based). Rule sanctions_2026-05-16.

    ▲ 1 escalation
  4. 04 REGEL

    HS code matching (47 line items)

    Line-item descriptions mapped against HS-code database. 42 line items deterministically unambiguous. 5 line items with unclear HS classification. Rule hs_code_2026_q2.

    ▲ 5 escalations
  5. 05 KI

    Language variant detection (EN, FIL, ES)

    Line-item texts in EN, FIL (Tagalog), ES detected. Model multilingual-bl-v3.1 classifies IMDG hints in non-English shipper instructions (Tagalog/Spanish pattern catches ~6-8 BLs/day that would otherwise mis-classify).

    Confidence 0.96 · threshold 0.85

    ✓ Classified
  6. 06 REGEL

    IMDG class matching (Amendment 42-24)

    39 line items non-hazardous. 8 line items IMDG class identified (Class 4.1 Lithium batteries x 6 - Standard UN3480, Class 9 Misc Dangerous x 2). VGM (SOLAS VI/2) cross-check against vessel stowage. Rule imdg_42-24_2026-01-01.

    ✓ 8 classifications
  7. 07 KI

    Dangerous-goods confidence routing

    6 lithium-battery line items with confidence > 0.90. 2 Misc-Dangerous with confidence 0.71. Model imdg-classifier-v4.2.

    2 line items confidence < 0.85 → escalation

    ▲ 2 escalations
  8. 08 MENSCH

    Customs case handler review (8 escalations)

    5 HS-code + 2 IMDG-confidence + 1 sanctions match. Case handler Anna B. (Customs team Speicherstadt). Top-3 suggestions per escalation pre-populated. Sanctions match resolved as false positive after Notify-party clarification.

    ✓ 8 decisions documented
  9. 09 REGEL

    EU-MRV + FuelEU Maritime + CII calculation

    Gross tonnage and voyage distance from vessel master. CO2 per EU-MRV methodology. FuelEU Maritime greenhouse-gas intensity per fuel mix (HFO/VLSFO/MGO blend). CII rating contribution recorded. Rule mrv_fueleu_cii_2026_q2.

    ✓ All 3 reportings ready
  10. 10 REGEL

    Audit-trail finalised

    All 9 decision steps with rule version, input hash, decider, confidence (AI steps), timestamp persisted. Hauptzollamt export message AES.E_USDA emitted to ATLAS via DAKOSY. Rule audit_trail_v1.4.

    ✓ Persisted + ATLAS submitted

Chapter 3 — Filipino crewing under MLC 2006 in English

Speicherstadt cluster - 5 minutes from engineering head office. English-language MLC compliance.

Head office Hallerstraße 8 is 5 minutes from Speicherstadt cluster. On-site meetings in HafenCity, Brooktorkai or Ballindamm reachable same day. Engineering counterpart sits in Hamburg, not Berlin or Munich. Workshop sessions in training centre at Grindelberg, 1.2 km from head office - separate room for works council sessions and auditor briefings without daily-business noise.

Filipino crewing under MLC 2006 as sister use case: Hamburg liner shipping companies (Hapag-Lloyd, MSC, Hamburg Süd, German tanker operators) employ around 70 percent Filipino crew under MLC 2006. The Manila bottleneck is real: POEA/DMW e-registration and OEC (Overseas Employment Certificate) issuance by the Department of Migrant Workers gate the entire crew-change cycle before any German visa question arises. Then SIRB (Seafarer's Identification & Record Book), PEME (Pre-Employment Medical Examination) validity windows, ILO 185 Seafarer's Identity Document, and the Schengen C-visa vs. transit visa interplay - the loose '72-hour window' is in practice a SID+transit-visa+OEC three-way race. MLC 2006 + IBF/ITF wage rates + flag-state regulations (Liberia, Marshall Islands, Antigua & Barbuda) apply on top. English-language workflow with Manila manning agencies (Hanseatic Lloyd, Wilhelmsen Ship Management, Magsaysay) standard.

Decision-Layer pattern identical to BL workflow: RULES validates MLC mandatory fields (sea-service months, PEME validity, STCW certificates, yellow card yellow fever, SIRB endorsements), wage calculation per IBF/ITF schema, POEA/DMW OEC pre-check, flag-state visa lookup. AI AUTONOMOUS handles multilingual document extraction (English + Tagalog + Spanish for Latin American crews), manning-agency contract validation, crew-change coordination between 3-7 ports. HUMAN escalation at wage-conflict (IBF vs. national agreement), PEME medical disqualification, or safety incidents in sea-service records.

Source code, prompts, rule sets and decision-records transfer to the shipping company by repository handover, contractually. After 12-18 months you operate your BL agent without us. Workshop at Grindelberg can include Filipino HR managers via remote bridge - or fully in-person if HR director from Manila is visiting Hamburg HQ. The success criterion is your exit from us, not the next retainer.

Related use cases

Frequently asked questions

What volume is realistic for a first Bills of Lading agent?
A model liner shipping company with 200 container vessels processes around 12,000 Bills of Lading daily (200 vessels x 60 BL/vessel/day). The first productive agent typically takes over the deterministic 70-80 percent (HS code matching, IMDG class, MRV reporting). Manual processing drops from 8-15 minutes per BL to under 30 seconds in the AI-autonomous path. Audit trail per line item: rule version, input hash, confidence score.
How is IMDG Code Amendment 42-24 handled?
IMDG Amendment 42-24 is mandatory since 1 January 2026. The Decision Layer integrates IMDG class mapping as a versioned rule (e.g. rule version `imdg_42-24_2026-01-01`). Class matching is deterministic in the rule-engine path. Ambiguous dangerous-goods classification escalates automatically to a customs case handler with top-3 suggestions. Audit trail documents per line item which IMDG version was applied.
How does the Filipino crewing workflow under MLC 2006 work in English?
It runs as a parallel use case driven by the 72-hour visa deadline after port call. MLC 2006 applies alongside IBF/ITF wage agreements and flag-state rules (Liberia, Marshall Islands, Antigua & Barbuda). The Decision Layer extracts structured data from boarding passes, sea-service records and manning-agency contracts in English and Tagalog, verifies it against MLC requirements, and generates visa applications with compliance evidence. Wage conflicts or missing sea-service months escalate to HR. Manila manning agencies get audit-trail visibility via API.
How does this Decision-Layer differ from classical RPA?
RPA maps fixed workflows. The Decision Layer separates each process step by decider type: HUMAN for discretion (dangerous-goods escalation), RULES if deterministic (HS code matching), AI AUTONOMOUS at confidence above threshold (language variants of B/L from 60+ countries). At low confidence, technically enforced human escalation - the agent cannot skip the case handler. Audit trail per step, unlike RPA black-boxes that only log start and end.
How fast is a first Bills of Lading agent productive?
4-6 weeks from first consultation to productive agent. The discovery workshop (1 week) analyses your current BL processing and maps the applicable rule sets - UCC, IMDG Amendment 42-24 and EU MRV as the core, with FuelEU Maritime, CII and sanctions screening layered on. The build phase (3-4 weeks) delivers a first productive pilot covering deterministic classification, the EORI/DAKOSY/ATLAS handshake, HBL/MBL distinction and full audit trail. Engineering is on-site at the Speicherstadt cluster, 5 minutes from head office Hallerstraße 8. The workshop runs in English with Filipino HR managers via remote bridge.
Does the architecture handle FuelEU Maritime and CII alongside EU-MRV?
Yes - EU-MRV alone is 2024-thinking. As of 2026, FuelEU Maritime (greenhouse-gas intensity per fuel mix: HFO, VLSFO, MGO, biofuels, methanol blends) and CII (Carbon Intensity Indicator vessel rating) run in parallel to MRV. The Decision Layer calculates all three per voyage from the same vessel-master input. Rule versioning carries the regulatory phase-in dates: <code>mrv_fueleu_cii_2026_q2</code>. Per-line-item CO2 attribution to BL is available for shipper-facing emissions reporting and Scope 3 inventories.
What about POEA/DMW + OEC + SIRB + ILO 185 SID on the Manila side?
These are the actual gating items, not the German visa. POEA/DMW e-registration plus OEC (Overseas Employment Certificate) issuance by the Department of Migrant Workers must clear before the German consulate processes a Schengen C-visa or transit visa for the crew change. SIRB (Seafarer's Identification & Record Book) endorsements, PEME (Pre-Employment Medical Examination) validity windows, ILO 185 Seafarer's Identity Document all sit upstream of the European leg. The Decision Layer pulls these via API where available (Hanseatic Lloyd, Wilhelmsen Ship Management, Magsaysay manning agencies) and escalates to HR if any pre-Manila document is missing or stale.

Schedule workshop at Grindelberg

3-day discovery: Day 1 process analysis, Day 2 Decision-Layer mapping, Day 3 use-case prioritisation. Concrete deliverable.

Schedule meeting

Discovery workshop below EUR 10,000. Pilot fixed price discussed after the workshop.

← Back to overview: AI Agents Hamburg (7 use cases)