Adani Ports · port-AI

Engagement summary

IndustryContainer terminal operations
RegionIndia (with Middle East rollout in flight)
Engagement typeMulti-agent deployment, TOS-integrated, multilingual
StatusLive in CFS Mag
ChannelsVoice, email, WhatsApp

This is the most mature multi-agent deployment Shipsy has live. Reference point for any port, terminal, or freight customer.

The business problem

Container terminal operations generate massive query volume — truckers, brokers, freight forwarders, customs agents all calling/emailing about the same handful of things: appointment status, container location, demurrage charges, regulatory holds. Five channels in, eight disconnected source systems behind them. Human operators stitched the experience together at huge cost.

The solution we delivered

Agents live in production

AgentWhat it does
Appointment SchedulingBooks, confirms, reshuffles dock slots
Customs AgentPre-Bayan filing, HS code validation, BoE / BoL / DO / LEO / Shipping Bill checks
Demurrage PreventionOutbound voice alerts before D-day charges hit
Document IntelligenceLLM-based extraction & validation across customs & trade docs
Invoice ReconciliationAuto-matches invoices vs. contractual tariff + SAP postings
Multilingual NLUEN · HI · GU · MR · AR — same agent across 5 languages

Volume distribution

The top 3 use-case categories are ~75% of all volume:

Category% of volume
Appointments & scheduling30%
Container & yard operations25%
Fees, demurrage & billing20%
Customs & regulatory12%
Constraints & cutoffs7%
Gate & access management6%

And ~60% of all volume is just three read-only lookups: appointment status, container availability, demurrage status. These are exactly what agents handle best.

Why this is the reference architecture

Federated stack today → orchestrated platform tomorrow, without replacing any source system. Shipsy slots between the channels and the 8 systems of record. One integration layer; six agents on top.

This is the pattern to copy for any port, terminal, or freight customer.

Implementation notes

Placeholder — needs input from Adani account team:

  • Multilingual rollout sequencing
  • TOS integration depth
  • Operator change management
  • Arabic-language rollout for Middle East expansion

Outcomes

Placeholder — needs numbers from Adani account team:

  • % of inbound volume resolved without human handoff
  • Mean handle time reduction
  • $ saved in demurrage prevented
  • CSAT change

Materials

  • Carrix proposal deck (uses Adani as the proof point) — Drive link
  • TOS integration architecture doc — TBA

Changelog

  • 26 May 2026: Initial draft from Carrix proposal deck (which features Adani as the live reference).