"Tofu's Efficiency Is Much Better": How GBS Malaysia Transformed Their Bookkeeping Workflow

10 branches, 500 clients, and a familiar problem

GBS (Global Business Services) is a division of Cheng & Co, one of Malaysia's established accounting networks. From their Kuala Lumpur headquarters, they coordinate 10 branches serving 400-500 clients across manufacturing, trading, e-commerce, and services. Their 48-person team handles outsourced accounting, payroll, and recruitment for clients in Malaysia, Australia, and Singapore.

That geographic spread means documents arrive in every format and language imaginable.

Mainland China invoices nobody could read

Chinese customers send documents in Chinese. Without OCR support for those characters, someone on the team had to translate before any data entry could begin.

Clients who combine everything into one PDF

Multiple invoices from different suppliers, all in a single file. The team had to manually separate them before processing could start.

Hard copies that won't go away

Despite encouraging clients to send soft copies, some still arrive by paper.

"If we have to capture manually? It's very time consuming."
— Michele Lim, GBS Assistant Director

They tried every OCR tool. None of them worked.

GBS wasn't new to automation. They'd been on Xero for four to five years and had experimented with multiple OCR solutions.

Problems with Other OCR Tools 1): accuracy wasn't there

Clients uploaded documents directly to an OCR solution. The problem?

"The accuracy is not 100%, I would say. It's not there."
— Michele Lim

The team couldn't rely on it. Many staff reverted to manual entry because they didn't trust the output.

Problems with Other OCR Tools 2): too complicated

Before any extraction could happen, they had to configure extensive rules for each client.

"There's a lot of functions that we need to set up before we using other tools. So nobody like to use. It's very complicated really."
— Michele Lim

The result: staff preferred typing

When your OCR tools require more effort than manual entry, people stop using them. That's exactly what happened.

"The accuracy is not there and very different. It's very difficult for us to keep on relying on the software. That's why some of the team members prefer using the manual way instead of just using other OCR systems."
— Michele Lim

The demo that actually impressed them

When Tofu was presented to GBS, Michele was cautious. They'd been burned before.

But something was different this time.

"We're quite impressed with their accuracy then that's why I decided to use one of the customers to test run. But quite surprisingly my team member like to use it."
— Michele Lim

That last part is worth reading again.

From skeptical to sold

The team needed warming up. After their experience with other OCR tools, they were understandably cautious.

"I think it's quite normal that the human being is like they need to try to adapt another solutions. As far as they already are quite adaptive, they will actually trust on these solutions."
— Michele Lim

The difference? Tofu actually worked. And the support made implementation smooth.

"The Tofu team has been giving us a lot of support. They even catch up with us every week to asking us whether we need help or not. They're always there with us. No matter how busy we are, they always be with us together."
— Michele Lim

4-5 hours to less than one hour

The numbers told the story immediately.

"They're even telling me last time they have to spend about like four to five hours just to key all the bills. But with the Tofu, I think they're telling me maybe less than an hour."
— Michele Lim

That's an 80% reduction in processing time. Per client. Per month.

But it wasn't just about speed. The entire timeline compressed.

"Let's say they have to complete the account they need about like one week to complete, but with this Tofu actually they can complete within just few days or just few hours only."
— Michele Lim

The AI that learns from corrections

Unlike rule-based systems that break when anything changes, Tofu gets smarter over time.

"If let's say maybe the first time they are doing wrong, so maybe the second time we just correct it back and then the third time they can make it correct already. The AI can actually learn what we actually do."
— Michele Lim

No rules to write. No templates to configure. The system learns from how you work.

No more overtime

Before Tofu, closing the books meant staying late. Month-end was a marathon.

Now?

"They can save more time and then maybe they can go back on time. No need to always spend on the overtime to just close the bookkeepings."
— Michele Lim

For a 48-person team spread across 10 branches, that's not just efficiency. That's quality of life.

From bookkeeper to advisor

The transformation went deeper than time savings. The nature of the work itself changed.

"Because now, I think basically they can change their roles from the bookkeeper to now as advisor roles."
— Michele Lim

Staff who used to spend their days on data entry are now doing higher-value work: reviewing, verifying, and advising clients.

"They can spend more time to do other value added things for our customers instead of just doing the double entries only."
— Michele Lim

Ready to take on more clients

When your team is drowning in manual work, growth means hiring. With Tofu, the math changes.

"Definitely, definitely. Because they can even do it faster and then again get more customer under their portfolios."
— Michele Lim

Same team. More capacity. No hiring required.

"They can serve bigger pool of the customers because I think traditional way is they have to key one by one. It's very time consuming for them."
— Michele Lim

Multi-receipt cropping: another 30-40% saved

During the interview, Michele learned about Tofu's multi-receipt cropping feature. Clients often paste multiple receipts onto a single page for expense claims.

Her reaction?

"This is very helpful to us because this one is definitely about like 30 to 40% of our time really is on this. Because I mean a lot of the expenses and transfer actually from their claims and we have to prepare the monthly claim just based on the customer receipts."
— Michele Lim

Another massive time sink, solved.

"Easy to use and save all the time"

When asked to describe Tofu in one sentence, Michele didn't hesitate:

"Easy to use and save all the time. Convenient."
— Michele Lim

For a team that had rejected every other OCR tool as too complicated or too unreliable, that's the highest endorsement.

What Michele would tell other firms

"I would suggest them to try. Because why not? You can reduce the team members' time and then they can do other better work instead of just normal double entries work."
— Michele Lim

The team that was skeptical about automation? They're now advocates.

The staff who preferred manual entry? They're requesting to use Tofu on more clients.

And the overtime that used to define month-end? It's becoming a thing of the past.

"Their efficiency is much more better than last time already. They impressed us."
— Michele Lim