How Phoenix Malaysia Saves 50%+ on Bookkeeping Time Across 140 Clients

From full manual to AI-powered: a skeptic's journey

When Pin Mei joined Phoenix Malaysia four years ago as Accounting Manager, the firm's entire bookkeeping operation ran on one system: manual data entry. Every invoice, every receipt, every transaction typed by hand. No AI. No automation tools. Just people and keyboards.

"When I joined four years ago, everything is manual."
— Pin Mei, Accounting Manager, Phoenix Malaysia

At the time, Phoenix managed around 30+ bookkeeping clients with a team of 7 to 8 preparers. Fast forward to today and the firm has nearly tripled its client base to over 140, with around 50 requiring monthly bookkeeping. The team has grown to 10 to 12 preparers. The math was becoming impossible.

The skepticism that almost stopped them

Pin Mei had heard about AI accounting tools before Tofu. The industry buzz had been building for years. But every conversation with colleagues pointed in the same direction: it does not work.

"A few years back, before we knew about Tofu, we heard about AI accounting. But at that time it was very premature. Nobody was really using it. Nobody knew how to use it."
— Pin Mei, Accounting Manager, Phoenix Malaysia

The skepticism ran deep and it was not irrational. Bookkeeping requires judgment. How do you teach AI to allocate costs across departments? How do you get software to understand that the same supplier invoice might need to be split three ways depending on the project?

"We were very skeptical. Like it's not going to work. Friends and colleagues all said the same thing."
— Pin Mei, Accounting Manager, Phoenix Malaysia

But the volume kept growing. More clients. More invoices. More transactions. The manual approach that worked at 60 clients was buckling under the weight of 140.

Why they gave Tofu a shot

The decision to try Tofu was not driven by excitement about AI. It was driven by necessity. Phoenix's parent organization put out a call for teams interested in testing AI bookkeeping tools. Pin Mei raised her hand.

"Things are getting improved. All the advancements. Then yeah, why not? If the AI can help us to do all the routine work, then why not try?"
— Pin Mei, Accounting Manager, Phoenix Malaysia

She started small. A handful of clients. First on Xero, then expanding to QuickBooks Online once the integration was available. Today, 14 clients run through Tofu, including some of the firm's most complex and transaction-heavy accounts.

How it actually works day to day

Pin Mei's clients range from straightforward trading companies to complex manufacturing operations. The straightforward ones were where Tofu proved itself fastest.

"For trading companies, F&B companies, it's pretty straightforward. Accounts payable just goes to cost of goods, expenses. As long as you manage to do the knowledge base correctly, it should be no problem."
— Pin Mei, Accounting Manager, Phoenix Malaysia

One food and beverage client sends hard copy invoices every month. The bookkeeper scans everything, uploads the batch to Tofu, and it captures the data correctly. Every month, the same suppliers. Vegetables, meat, supplies. The AI learned the patterns within weeks.

"After one month, the second month is already better. The invoices are almost the same every month. Buying the vegetables, buying the meat. So it's pretty fast."
— Pin Mei, Accounting Manager, Phoenix Malaysia

The manufacturing clients are more complex. Some suppliers provide materials used across multiple departments, meaning a single invoice might need to be split across three different account codes. This is where the human judgment still matters, and where Pin Mei sees the AI and accountant working together rather than the AI replacing anything.

"Sometimes one supplier, they are supplying one item. But this item can be used for the operation, can be used for admin. So the AI won't be able to know which department. You might need manual adjustments sometimes."
— Pin Mei, Accounting Manager, Phoenix Malaysia

But even on these complex accounts, Tofu handles the extraction, the data capture, the initial coding. The bookkeeper's job shifts from typing to reviewing and adjusting.

The numbers: more than 50% time savings

Pin Mei does not sugarcoat the results or inflate them. When asked how much time Tofu saves on the clients that are running well, her answer is matter of fact.

"For my cases, I will say it's more than half. More than 50% of my time."
— Pin Mei, Accounting Manager, Phoenix Malaysia

The time savings compound as the AI learns. The first month requires corrections and feedback. By the second month, the system is noticeably more accurate. By the third or fourth month on complex accounts, it stabilizes.

Getting the team on board

Implementing AI in a 16-person team is not the same as implementing it for yourself. Some team members were resistant. Some were fearful. Pin Mei's approach was pragmatic rather than evangelical.

"The world changes. We need to improve ourselves. You cannot just be doing the same old thing, repeat and repeat. If there is another tool that can help you, why not try?"
— Pin Mei, Accounting Manager, Phoenix Malaysia

She draws a parallel to ChatGPT and other AI tools her team already uses daily. If they are comfortable using AI for general tasks, the leap to AI for accounting is smaller than it feels.

"Since you are using ChatGPT, why not add another AI? AI Accounting."
— Pin Mei, Accounting Manager, Phoenix Malaysia

This year, Pin Mei plans to onboard junior staff onto Tofu, expanding beyond the senior team members who have been using it so far. She expects the platform learning curve to be minimal — just a couple of days.

"Definitely worth it"

When asked whether the effort of rolling out Tofu across a large team is worth it, Pin Mei does not hesitate.

"Definitely worth it. Because once you get it trained, basically you don't have to worry for the rest. Unless there is a big change, you don't have to worry about the remaining days."
— Pin Mei, Accounting Manager, Phoenix Malaysia

This is the key insight from a firm that has been live for a full year. The setup period requires investment. The knowledge base needs feeding. But once it stabilizes, the ongoing maintenance is minimal. The AI remembers what it has learned. New staff benefit from the corrections that previous team members made.

Support that keeps up

Pin Mei is known internally at Tofu for her detailed feedback. She screenshots issues, annotates them in Excel, and sends them to the support team with surgical precision.

"Super fast. Super fast. I am the one who is very slow in responding."
— Pin Mei, Accounting Manager, Phoenix Malaysia

She credits the support team's speed with enabling her to push through the inevitable bumps that come with any AI implementation.

The bigger vision: from data entry to advisory

For Pin Mei, the point of automation is not about doing the same work faster. It is about fundamentally changing what her team spends their time on.

"Hopefully with all these AI tools, it can help us to replace all the routine work and let them focus more on the advisory. Advisory you cannot replace by AI because you need judgment, you need analytical skill."
— Pin Mei, Accounting Manager, Phoenix Malaysia

She is direct with her team about what AI means for their roles. It is not a threat. It is a shift.

"It is not going to replace you because at the end of the day it will still need your judgment. What the client is looking from us is our advice. How we can help them, what are the potential problems, what are the potential growth areas. Those are the valuable things."
— Pin Mei, Accounting Manager, Phoenix Malaysia

What comes next

Phoenix Malaysia is expanding its Tofu usage in 2026.

More clients being onboarded. Junior staff being trained on the platform. New document types like accounts receivable invoices and direct expenses being tested alongside the core accounts payable workflow.

The firm went from 70 clients to 140 in three years. The accounting team grew from 10 to 16. Without automation, that growth equation would have required far more headcount. With Tofu handling the extraction and initial coding, each team member's capacity stretches further.

When asked to give one piece of advice to firms considering AI bookkeeping tools, Pin Mei's answer reflects someone who started as a skeptic and became a champion through lived experience rather than hype.

"You must constantly give feedback. What are the issues you face? What are the problems? Then the team knows what we are facing. Then they do whatever they can to make sure it fits."
— Pin Mei, Accounting Manager, Phoenix Malaysia

It is not magic.

It is a partnership between the accountant's expertise and the AI's ability to learn. Phoenix Malaysia invested the time to teach the system, and a year later, the system is teaching their newest hires.