Logistics

Pilot wave 2 (M7). Impact: high. See the hub.

Current state (assumptions to validate)

Logistics turns confirmed orders into deliveries at the berth: planning the refrigerated and dry fleet across the day’s vessel calls in Piraeus and nearby ports, coordinating with agents on berth/anchorage access and launch services, and — the heaviest part — producing customs and bonded documentation for ship supplies (export-equivalent supply declarations, bonded-warehouse removal documents, transit paperwork for items moving between sites). Vessel schedules shift constantly; a delayed berth or an advanced ETD reshuffles the whole day.

Assumptions — to validate in discovery: team ~5–8 plus drivers; planning on whiteboard/Excel + phone; customs docs drafted manually per shipment against templates, filed on paper; ETA/ETD updates arrive by email/phone from agents; cold-chain documentation (temperature at loading/delivery) manual. Pain points: constant re-planning under schedule churn, customs-document drafting time and error risk (rejections cost hours against a sailing), no consolidated view of the day’s deliveries vs deadlines, proof-of-delivery paperwork returning slowly.

Target AI-first operating model

By M12 the day’s delivery board is generated each morning and re-ranked automatically when ETAs change: an agent parses agent/port emails for schedule updates and flags affected deliveries. Customs and bonded documents are pre-drafted from the SoftOne order and shipment data — the documentation officer reviews and lodges, never types from scratch. Drivers confirm deliveries with photographed PODs that file themselves. Humans own all customs lodgings, route decisions, and customer/agent coordination calls.

AI use cases

Use case	Pain addressed	Data needed	Complexity	Impact	Pilot
Customs/bonded document pre-drafting from order data	Manual drafting, rejection risk, deadline pressure	SoftOne orders/items (HS codes, values), doc templates, bonded records	H	H	Y
ETA/ETD change detection from agent emails → delivery-plan re-flag	Schedule churn handled by memory	Logistics mailbox, delivery plan, order deadlines	M	H	Y
Daily delivery-board generation (deliveries ranked by sailing deadline, cold-chain flags)	No consolidated view	SoftOne orders, fleet availability, warehouse readiness	M	H	N
POD capture & filing (photo → order match → archive)	Slow paperwork return, disputes	Driver photos, SoftOne delivery records	L	M	N
Route/load suggestion for multi-drop berth runs	Improvised routing	Delivery board, vehicle constraints (reefer/dry)	M	M	N
Customs-rejection pattern analysis	Repeat errors	Historical docs and rejection reasons	L	M	N

Process transformation opportunities

• Documentation flow: from per-shipment manual drafting to review-and-lodge on pre-drafted documents, with a rejection-feedback loop.
• Schedule intake: from phone/email memory to parsed, logged ETA/ETD changes driving the plan.
• POD lifecycle: digital capture at the berth, auto-filing, same-day availability to finance for invoicing.
• Delivery board: one live view shared with warehouse and CS, replacing status phone calls.

Required data sources

• SoftOne: orders, delivery data, item customs attributes (HS codes — completeness assumption, to validate), bonded stock records.
• Logistics mailbox (agent correspondence, port schedules); customs document templates and past filings (paper → scan).
• Fleet list with reefer/dry capabilities; POD archive.

Potential AI agents

• Customs Document Drafting Agent — pre-drafts supply/bonded/transit documents from order data; documentation officer reviews and lodges every document.
• Schedule Watch Agent — parses schedule-change emails and flags affected deliveries; dispatcher approves plan changes.
• POD Filing Agent — matches delivery photos to orders and files them; spot-check by team lead.

Automation opportunities

• n8n: morning delivery-board generation → Teams post to logistics/warehouse/CS.
• POD photo intake flow (driver phone → SharePoint → match).
• Escalation alert when a sailing-critical delivery lacks documents N hours before ETD.

Required integrations

• SoftOne read (orders, deliveries, item customs data) — see integrations.
• M365 Outlook (logistics mailbox), SharePoint (doc archive), Teams.
• Customs e-filing system (lodging stays human via the official channel; no direct integration in year 1).

KPIs

KPI	Baseline	M12 target
Customs-doc preparation time per shipment	TBD (M2)	−60%
Customs document rejections	TBD	−70%
Deliveries late to berth (logistics-attributable)	TBD	~0
PODs filed same day	TBD	≥95%
Schedule changes captured in plan (vs missed)	TBD	≥98%

Risks

• Erroneous pre-drafted customs document lodged unchecked → mandatory human review with field-by-field source display; officer accountability unchanged.
• Bonded-stock documentation error creates customs liability → bonded data always from SoftOne records, never inferred; audit trail on every draft.
• Over-reliance on parsed ETAs → schedule flags are advisory; dispatcher confirms with agent for critical calls.

Training needs

• Documentation officers: draft-review workflow, source verification (M6–M7).
• Dispatchers: delivery board, schedule-watch flags (M7).
• Drivers: POD capture app habit (M9).

Deliverables

• Customs-document template digitization + drafting pilot (M7), production (M10).
• Live delivery board; Schedule Watch Agent; POD pipeline.
• Logistics KPI dashboard.

12-month execution milestones

Month	Milestone
M1–M2	Discovery: document-type inventory, delivery process mapping, baselines
M3	Customs templates digitized; HS-code data quality assessed
M4	Assistant onboarding
M6	Offline accuracy testing of document drafts; officer training
M7	Wave-2 pilot: customs-doc pre-drafting + schedule watch on live shipments
M8	Pilot evaluation; delivery board live
M9–M10	Agents in production; POD pipeline rolled out to drivers
M12	KPI review; handover