Mate in the Loopby Kenny Jung
KOEN

AI agent · AI / MLMaritime & ports

DatamarNews reported on October 15, 2026 that CMA CGM and HappyRobot are scaling AI voice agents across global operations after pilots in sales, customer service and finance.

AI-drafted · reviewed by KennyCodex check · partly supported2 min read

Source [1] datamarnews.com · 15 Oct 2026 · “AI Voice Agents at CMA CGM Expand Across Global Operations with HappyRobot” ↗
Adoption stage: Production — A named organisation runs it in day-to-day operations.

Summary

DatamarNews reported on October 15, 2026 that CMA CGM and HappyRobot are scaling AI voice agents across global operations after pilots in sales, customer service and finance. The sales agent "Eva" reached over 4,000 customers in UK and Senegal campaigns, about a third of whom moved into active shipping discussions; in customer service the agents handle cargo-release requests — status checks, customs clearance, documentation — and resolve 46% of customer-service requests directly; in finance they chase outstanding invoices. CMA CGM, which handles 54 million customer requests a year, runs the agents in Europe and North America today and plans to expand to Africa, South America and Asia, with Eva in 20 countries by the end of 2026.

Basis: the source [1], facts F1–F5

Kenny's take

What changes at the carrier's customer desk is who handles cargo-release queries, quotes and invoice follow-up. In Europe and North America this is production, and the agents close 46% of customer-service requests on their own. How and when the rest reach a person is not in this article — that is the next thing to check. The take weakens if Eva is not in 20 countries by the end of 2026, or if later reporting puts the 46% lower.

AI draft · approved by Kenny

Analysis

Stage: production. A named carrier reports the agents as "currently operational in Europe and North America" after completed pilots, with a measured resolution share (46%) in customer service. The 20-country target for Eva is a plan.

For customer-service-voice-agents in maritime, the work moved to the agent is customer-facing — quotes, cargo release, invoice follow-up. For q-agent-ops-hitl, the source gives the 46% directly resolved but, as fetched, does not describe how the remaining requests are escalated to people or what approval the agent needs before acting; that is the missing evidence.

Facts from the source

  1. F1CMA CGM Group and HappyRobot "are scaling AI voice agent deployment across global operations following successful pilots in sales, customer service, and finance functions" (DatamarNews, October 15, 2026). [1]
  2. F2Sales: the AI agent "Eva" contacts prospective customers to understand shipping needs and provide freight quotes; Eva "contacted over 4,000 customers in UK and Senegal campaigns; approximately one-third engaged in active shipping discussions". [1]
  3. F3Customer service: the agents handle cargo release requests, "including status checks, customs clearance, and documentation"; "46% of customer service requests resolved directly by AI agents". [1]
  4. F4Finance: the agents support payment follow-up by contacting customers about outstanding invoices. [1]
  5. F5CMA CGM "processes 54 million customer requests annually"; the deployment is "currently operational in Europe and North America", with expansion planned for Africa, South America and Asia and Eva targeting 20 countries by the end of 2026. [1]

Organisations: CMA CGM, HappyRobot

Counterpoint — verification notes

Codex checked each sentence against the source and searched for counter-evidence. Verdict: Codex check · partly supported.

  • Other sideCMA CGM’s own announcement specifies more than 46% of cargo-release enquiries and more than 54 million annual customer requests. It confirms the named carrier’s involvement. [2]
  • Other sideAnother outlet likewise limits the resolution figure to cargo-release enquiries and notes that the measurement period and request volume were not disclosed. [3]
11 sentence-level checks
  • factspartly-supported: AI agents directly resolved 46% of customer service requests. — The article says more than 46% of cargo-release requests, not 46% of all customer service requests.
  • factspartly-supported: CMA CGM processes 54 million customer requests annually; agents are operational in Europe and North America, with expansion planned elsewhere and Eva targeting 20 countries by end-2026. — The article says more than 54 million, and describes the regional rollout as being implemented rather than fully operational. The expansion and Eva’s 20-country target are supported.
  • summary_enpartly-supported: The agents resolve 46% of customer-service requests directly. — “These requests” refers to cargo-release requests; the article also says more than 46%.
  • summary_enpartly-supported: CMA CGM handles 54 million customer requests a year. — The source says more than 54 million.
  • summary_enpartly-supported: CMA CGM runs the agents in Europe and North America today. — This states a firmer regional operating status than the article.
  • summary_kopartly-supported: 고객 서비스 요청의 46%를 에이전트가 직접 해결했다. — The percentage is more than 46% and applies to cargo-release requests, not all customer-service requests.
  • summary_kopartly-supported: CMA CGM은 연간 고객 요청 5,400만 건을 처리한다. — The source says more than 54 million.
  • summary_kopartly-supported: 현재 유럽·북미에서 운영 중이다. — The article describes implementation in those regions, a weaker status than the summary states.
  • analysispartly-supported: The agents are currently operational in Europe and North America, with a 46% customer-service resolution share. — The regional status is overstated; the measured share is more than 46% of cargo-release requests.
  • takepartly-supported: In Europe and North America this is production, and agents close 46% of customer-service requests on their own. — Production is a defensible overall stage, but the article does not establish that status across both regions; the percentage applies only to cargo-release requests and exceeds 46%.
  • takepartly-supported: 유럽·북미에서는 운영 적용이고, 고객 서비스 요청의 46%를 에이전트가 직접 해결한다. — The regional operating status is stronger than the article’s wording, and the metric is more than 46% of cargo-release requests.

Sources

  1. [1] datamarnews.com — “AI Voice Agents at CMA CGM Expand Across Global Operations with HappyRobot” original
  2. [2] cmacgm-group.com counter-evidence (Codex)
  3. [3] modaisemfoco.com.br counter-evidence (Codex)

Keep reading

Same industry

en.sedaily.com

HD Hyundai Heavy Industries has built AI CIP (AI Coating InsPector), a vision-AI and deep-learning system that finds and analyses coating defects on ships, and on October 29 signed an MOU in Ulsan with four classification societies — LR, ABS, DNV and KR — joined by HD Korea Shipbuilding & Offshore Engineering and HD Hyundai Samho.

The system has been developed with HD KSOE's AX Center since 2024 on coating quality data from the yards, replacing inspections that depended on workers' eyes and experience and varied with inspector skill. Under the MOU the two yards expand its use and the four class societies work on technical reviews and standard inspection procedures.

Maritime & portsAI / MLStage · Pilot
Codex check · partly supportedRead the take
The Maritime Executive

Cisco and PSA Singapore will build a proof of concept of an AI-powered observability and incident management platform at Tuas Port, which the release calls the world's largest fully automated container terminal.

The platform pulls real-time performance data from PSA's applications, infrastructure and network into one view and uses AI to map service interactions so that, during an incident, technical teams can see which services are affected, diagnose the root cause and get resolution recommendations. It is meant to reduce and eventually pre-empt outages such as Terminal Operating System (TOS) failures, starts at the Tuas Port Phase 2 Living Lab, and is justified with Cisco research putting average annual downtime cost per organisation at US$300 million.

Maritime & portsAI / MLStage · Pilot
Codex check · partly supportedRead the take
쉬핑뉴스넷

HD Hyundai and Anduril ran a 3.5-hour autonomous-operations demo on a converted 368-ton US Navy test vessel off Long Beach, with the Navy, Coast Guard and ABS watching.

Beyond navigation and formation sailing, it showed an engine-failure emergency response driven by an engine-autonomy layer that monitors machinery and reacts on its own. It is a bounded demonstration on a defence platform, not commercial service.

Maritime & portsAI / MLAI agentStage · Pilot
Codex check · partly supportedRead the take

Same AI technology

전기신문

At the Korean Society for Railway autumn conference on October 29, Lee Dong-seop of Daeati presented the AI train-conflict prediction and resolution system planned for the Second Railway Traffic Control Center (about KRW 250 billion, completion targeted for 2028).

It couples physics-based micro-simulation (gradient, curves, track length, vehicle specs) with AI trained on actual operating records and track occupancy, aiming to predict about 2 hours ahead and compute conflicts and resolution options in about 3 seconds; resolution runs reinforcement learning first, which handles about 995 of 1,000 cases, with a CP-SAT solver for the rest. The system presents conflict types and options and the dispatcher makes the final choice on punctuality, cost and energy; the Guro control center is to be upgraded to the same system for redundancy.

Other industriesAI / MLData science & classical MLStage · Research
Codex check · partly supportedRead the take
engineering.com

Siemens and P&G are rolling their jointly developed Visual Inspection Cockpit (VIC) out across P&G's manufacturing operations worldwide.

It inspects products in real time at full line speed; depending on the product, scrap rates are down 10 to 20 percent, and new deployments commission five to ten times faster than traditional custom vision systems. P&G's deep-learning models run near the equipment on Siemens Industrial Edge with Nvidia-GPU industrial PCs, and the system can automatically trigger alerts or removal of defective products from the line.

ManufacturingAI / MLStage · Production
Codex check · partly supportedRead the take
SecurityWeek

Cisco has open-sourced Project Quarry, a neurosymbolic rule language for detecting AI-agent attacks: rules combine neural evidence from ML classifiers and LLM judges with symbolic checks on conditions, sequences, string matches and tool relationships, across single- and multi-turn LLM conversations, agents, tool use, MCP interactions and multimodal inputs, for runtime guardrails, post-hoc analytics and incident response.

The article cites 2025 campaigns in which Claude Code was used against around 30 organisations and against Mexican government institutions, split into small routine-looking tasks. Cisco wants Quarry to become an open standard like Snort, YARA and Sigma, and points to Hugging Face reconstructing roughly 17,600 agent actions after the fact as the kind of sequence it aims to catch in progress.

Other industriesAI securityAI agentStage · Research
Codex check · partly supportedRead the take

Same stage, another industry — the pattern elsewhere

chrobinson.com

On October 26, 2026 C.H.

Robinson said AI agents now do work in its ocean, air and customs-brokerage operations. For U.S. Customs, an agent extracts data from more than four dozen document types arriving from shippers, carriers and government agencies while the system flags and requests missing ones; time spent fell 23% and customs brokers now concentrate on validation and compliance. Quoting agents have produced over 80,000 quotes for 3,600+ customers, cutting turnaround from 1–3 days to 2–3 minutes, and ocean-booking agents for the thousands of weekly booking requests went live in China in August, with the rest of Asia in Q4 and global availability expected in Q1 2027 — after about 10 years of earlier digitisation reached only about 5% automation of ocean and air booking requests.

Logistics & supply chainAI agentAI / MLStage · Production
Codex check · partly supportedRead the take
forever.security

Forever Security (Gal Weizman and Omer Samet, October 22) reports a live attack in which a website silently drops a malicious AGENTS.md or CLAUDE.md into ~/Downloads and any coding agent started there — OpenAI Codex or Claude Code — follows it as trusted instructions.

Against Codex the file registers a malicious MCP server in ~/.codex/config.toml that installs the known-malware npm package function-color, with over 90% success in testing; against Claude it redirects API traffic through ANTHROPIC_BASE_URL to an attacker domain for OAuth token theft and code execution, at about 50% initial compromise and about 20% same-session execution. The firm says there is nothing for OpenAI or Anthropic to patch and recommends never starting agents in Downloads, auditing instruction and config files, and keeping approval gates and sandboxes on.

Other industriesAI securityAI agentLLMStage · Production
Codex check · partly supportedRead the take