Hernan halves order entry time and makes its durian cold chain traceable end to end with Claude
A durian doesn’t wait for a spreadsheet. From the moment a Musang King fruit is graded at the receiving dock to the moment it leaves, still cold, for one of eighteen export markets, every handoff that slows down is quality the fruit doesn’t get back.
Hernan Corporation Sdn Bhd is a Malaysian exporter of frozen and fresh durians and durian-based products, and one of the country’s leading names in durian cold-chain logistics. A 49-person operation supplies eighteen export markets through a chain of time-critical stages — sales, pre-shipment documentation, grading and receiving, pre-cooling, final packaging, driver assignment, inspection, loading and delivery — each run by a different team, each handing the fruit and its paperwork to the next.
The challenge
Hernan’s problem was never demand. It was that the operation left no trail. The prior process ran on paper forms, Excel sheets and forwarded WhatsApp messages: a record could be lost between the dock and the system, an order’s status lived in whichever chat the update happened to land in, and data reached management hours after the fact.
The cost of that showed up at every handoff. A sales order meant three minutes of Excel work followed by manual key-in of every line item, with the file forwarded over chat and no traceable record left behind. At grading and receiving, a staff member photographed the receipt and sent it to an admin, who downloaded it and keyed in every durian variety and weight by hand — no proper records, no traceability, and a delay that compounded through peak season. Pre-cooling and final packaging ran on handwritten forms. And because every member of staff recorded things slightly differently, the same data looked different depending on who wrote it down.
The approach
Deployed under Anthropic’s Claude Partner Network in Malaysia, the build divided the job cleanly between two systems. Lark is the record — one structured, traceable home for every process. Claude is the hands — auto-extracting data, auto-calculating, auto-standardising formats and updating the system the instant a photo or file is shared.
The agent, @Hernan Operation Assistant, sits inside the Lark group chats each team already used to coordinate. The interface the staff touch did not change; what happens behind it did. A sales team member uploads a completed order file from desktop or mobile and tags the agent, which summarises it, extracts the line items and updates the system automatically. At three points further down the chain — grading and receiving, pre-cooling and final packaging — a staff member simply photographs the fruit and tags the agent, which reads the image with Claude’s vision capabilities, asks any follow-up questions, and turns the photo into a structured inventory entry by durian grade and weight.
The moment a sales order is created, the agent notifies every downstream team at once: pre-shipment to upload shipping documents, pre-cooling to execute the order against its spec list, final packaging to stand by for incoming stock, and loading to assign a driver for delivery to KLIA. Every stage starts on a notification rather than a chase. Loading, inspection and delivery run on the same pattern with forms instead of photos, and two exception modules sit on top: a rejection form available at pre-cooling and final packaging, and an escalation channel that alerts automatically on inventory discrepancies, rejections and departure cut-off timing, with a standing reminder to every operational group an hour before a shipment’s scheduled departure.
Management sits above all of it with a live feed of Claude-generated operational insight in its own group — completion rates, delivery volume in box counts, and rejection and inventory breakdowns by grade, weight and reason — produced as a byproduct of the same workflow, with nobody compiling a report.
The results
The before-and-after is sharpest at the handoffs, and measurable at each one.
| Sales order | Before: three minutes of Excel work per order, then manual key-in of every item, with the file forwarded over chat and no traceable record. After: one upload from desktop or mobile — the agent summarises, extracts and updates the system automatically. Order entry time cut by 50%. |
|---|---|
| Grading & receiving | Before: a receipt photo sent to an admin over chat, who downloaded it and keyed in every durian variety and weight by hand — no records, no traceability, delay compounding in peak season. After: the photo goes straight to the agent, which extracts the data and updates the system immediately. The entire step completes within two minutes, fully traceable. |
| Pre-cooling & final packaging | Before: handwritten forms with no proper records — an order’s status lived in chat messages, not the system. After: photo-based capture from a phone on the floor gives both stages full traceability, with real-time updates feeding management dashboards. |
| Coordination | Before: each team chased the one before it for status. After: every stage starts on an automatic notification the moment the preceding stage completes — pre-shipment, pre-cooling, final packaging and loading are all told at once when a sales order is created. |
| Data consistency | Before: the same data looked different depending on which staff member recorded it. After: the agent, not the individual, defines the structure — formats are standardised automatically on capture. |
| Management visibility | Before: data reached management hours after the fact, and a full picture took a day of chasing. After: each stage stamps the record in real time and the dashboard reflects it immediately — visibility exists by default. |
| Scale | No headcount added. The same 49-person team supports eighteen export markets. |
What it unlocked
The point is not that Hernan replaced paperwork with a chatbot. It is that a business whose product loses value by the hour now has a live, stage-by-stage record of every order, from the sales desk to the delivery truck, without adding a single person to keep pace with eighteen export markets. Every order became a traceable journey instead of a trail of messages.
Adoption cost almost nothing in change management, because the capture surface is the one the staff were already using. Nobody was trained onto a new system. They upload a photo and reply to a message, exactly as before — and the extraction, the grading breakdown, the format standardisation and the escalation judgment happen behind it.
“Our durians don’t wait for paperwork, and now our systems don’t either. Claude lets our teams turn a photo into a verified inventory record and an order into a tracked shipment, in the same group chat they were already using — that’s the kind of speed a perishable export business has to run on to serve eighteen markets at once.”
— Datuk Anna Teo, Founder & Group CEO, Hernan Corporation