The verdict in three sentences
A Singapore freight forwarder receiving 1,500 emails a day across four shared inboxes (quotes, container tracking, customs documents) takes about 6 hours on average to send a first reply, and some quote requests simply get lost. An AI classification layer that reads every email, extracts the job number, bill of lading or container number, routes it to the right coordinator and drafts a reply costs in the range of 30,000 to 55,000 SGD, plus 150 to 400 USD a month in API fees. Delivered in 8 weeks, it targets a 1-hour first response, with human approval on everything that goes to a customer.
Shared-inbox chaos around the port of Singapore
Around Tanjong Pagar, Pasir Panjang and the new Tuas terminals, a mid-sized forwarder handles hundreds of open jobs: transshipment, LCL consolidation, air freight from Changi, cross-border trucking to Malaysia. Emails hit quotes@, ops@, customs@ and the general inbox, often with several people in copy. Every morning one or two staff spend two to three hours sorting, forwarding and hunting for the job a message belongs to when it only says "my container MSKU...".
| Metric | Manual triage today | With AI classification |
|---|---|---|
| Emails per day | 1,500 | 1,500 |
| Daily triage time | 5 to 6 h combined | under 1 h of review |
| Emails linked to the right job | 70 to 80 % | 95 % target |
| Average first response time | 6 h | 1 h target |
| Quote requests unanswered after 24 h | 10 to 15 % | under 2 % |
| Workload visibility per coordinator | none | real-time dashboard |
Lost quotes hurt most. A shipper without an answer by mid-morning sends the request to three competitors, and the first clear price often wins the shipment.
What the AI does, step by step
The system connects to Microsoft 365 or Google Workspace. For each incoming email it assigns a category (quote request, tracking query, document to process, claim, invoice, spam), extracts identifiers (internal job number, BL, ISO 6346 container number, TradeNet permit number), links the message to the job in the freight management system (CargoWise, Magaya or a local TMS) and assigns it to the responsible coordinator. For simple questions ("where is my container?") it drafts a reply using the real status from the system, which the coordinator approves in one click.
| Cost item (2026 ballpark, Singapore) | Range |
|---|---|
| Scoping, categories, routing rules | 4,000 to 7,000 SGD |
| Inbox and freight system connection | 8,000 to 16,000 SGD |
| AI classification and extraction engine | 10,000 to 18,000 SGD |
| Draft replies and dashboard | 5,000 to 10,000 SGD |
| Team training and acceptance testing | 3,000 to 4,000 SGD |
| Monthly API cost (1,500 emails a day) | 150 to 400 USD |
| Time to production | about 8 weeks |
Total budget therefore lands between 30,000 and 55,000 SGD. The most variable item is the freight system connection: CargoWise eAdaptor or a modern API connects quickly, an older local TMS needs an intermediate database or scheduled exports. Singapore SMEs can check whether the project qualifies for Enterprise Singapore support schemes, which can offset part of the cost.
Guardrails to insist on
Nothing reaches a customer without human approval for at least the first three months. Emails involving amounts, claims or customs documents always stay in manual review. The system must log every classification so errors can be corrected and rules improved weekly.
On data, Singapore's PDPA requires you to document the processing, pick an AI vendor that does not train on your emails, and limit retention. Many shippers also include data clauses in their contracts.
For a related use case, see our article on AI-written field service reports and our AI services.
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Wei Ling, operations director at a forwarder in Pasir Panjang, runs 14 coordinators and 1,500 emails a day. She funds a 42,000 SGD project with an API cost of 250 USD a month, about 335 SGD.
- Triage time saved: 5 hours a day, about 1,300 hours a year over 260 days.
- At 28 SGD per loaded hour for an ops executive, that is 36,400 SGD a year.
- Quotes recovered: out of 60 requests a day, going from 12 % to 2 % slow or missed saves 6 quotes a day. If 1 in 10 converts into a job worth 900 SGD margin, the gain is about 140,000 SGD a year.
- Even dividing that assumption by four, the project pays back in under 8 months, API costs included (about 4,000 SGD a year).
FAQ
Can the AI handle mixed English and Chinese emails?
Yes, current models handle mixed-language emails from carriers and shippers well. Correct classification usually exceeds 90 % after two weeks of tuning.
Can it read a container number inside a PDF attachment?
Yes, extraction also works on BLs and invoices in PDF, including scans, at a slightly higher API cost within the 150 to 400 USD monthly range.
Do replies go out automatically?
No, at launch every draft is approved by a coordinator. After three months, only simple status replies may be automated, if management decides so.
Do we need to change our email or freight system?
No. The system connects to Microsoft 365 or Google Workspace and to your existing system via API or export, within the 8-week timeline.
What happens to our customer data?
It stays in your mailboxes and your system; the AI vendor processes emails without retaining them for training, in line with PDPA obligations.
Let's scope your project. We define your email categories, routing rules and freight system connection for go-live in about 8 weeks. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
