Case studies: real automation, less manual entry — NoType
NoType / Case studies Rev. 2026 · JB, MY
Case studies

Real businesses.
A lot less typing.

A customs forwarder and a wholesale supplier — three manual jobs we took off their desks.

Customer 01

A JB customs forwarder

§ 01 / The problem

The problem

A busy customs forwarder receives hundreds of supplier invoices a day. Each one has to be re-keyed into the K2 declaration form by hand: vendor details, item lines, amounts, classification codes. The work is fragile — one typo costs a re-submission. During peak season, staff stay late just to clear the backlog.

§ 02 / The setup

The setup

We installed a folder watcher on the customer's PC. Drop an invoice PDF in, and AI extracts every field that K2 needs. The first batch from each new vendor goes through a quick review queue so any extraction quirks get caught before they become declarations. Once the data passes review, keyboard automation types it directly into the customer's existing K2 software — exactly the way a person would, just faster and without typos.

§ 03 / The numbers

The numbers

Invoices / day
0+

Invoices processed per day

Up from a manual ceiling of ~80.

Entry time
5 min 10 sec

Per-invoice entry time

A 0% reduction.

Accuracy
0%

Accuracy after review phase

Once a vendor's quirks are mapped, errors are vanishingly rare.

§ 04 / Endorsement

Customer endorsement coming soon.

— Customer name pending public approval
§ 05 / We didn't stop at K2.

The government upload, on autopilot.

Every finished declaration still had to be logged into a government portal and uploaded by hand — the same clicks, every time, all day. So we built that step too: the document gets where it needs to go without anyone babysitting a browser.

Built and proven. Same engine as the K2 flow.

Customer 02

A wholesale supplier

§ 06 / The problem

Orders live in a WhatsApp group. They shouldn't get lost there.

Orders land in a busy group in fast shorthand — customer, lorry, then a stack of lines like CN LATE LANE (48)-2CTN/RM73, half in Malay. Someone has to read every message and retype it into a clean list. Miss one while you're serving a customer, and that order just… doesn't get picked.

// This one sold fruit — but it's the same story for any supplier.

What we built: a bot that sits in the group, reads every order — the shorthand, the mixed languages, all of it — and turns each into a clean, structured order: item, quantity, size, price. Anything it can't read with confidence gets flagged for a human instead of silently dropped.

Distributor, retailer, supplier — if orders come in through a group chat, it's the same job.

Built and proven on real orders. Rolling out now.

fig.01 — messy WhatsApp order → structured order The chat is fake. The mess is real.
§ 07 / Next step

Want this for your business?

Tell me the manual work that's wasting your team's time. I'll tell you whether automation makes sense.

Chat on WhatsApp

// JB-based. I usually reply within an hour during work hours.