Last tax season nearly broke Margaret Liu’s practice. Three staff, 400 clients, and a mountain of paper receipts that seemed to grow faster than they could process it.

This year was different. Not because of more staff—she still has three. Because of a few carefully chosen AI tools that changed how work gets done.

“I was skeptical,” Margaret admitted. “I’ve been doing this for 22 years. I didn’t think software could understand the nuances of tax work.”

She was partly right—and partly wrong.

What Changed

Margaret’s Hurstville practice implemented three main changes:

Receipt processing: Instead of manually entering expense receipts, clients photograph them via an app. AI extracts the vendor, amount, date, and categorises the expense. Staff review rather than transcribe.

Document summarisation: Long financial statements from banks and superannuation funds get summarised by AI before staff review. Key figures are extracted automatically.

First-draft correspondence: Standard client letters—lodgement confirmations, information requests, deadline reminders—are drafted by AI based on templates and client data. Staff edit rather than write from scratch.

The Numbers

Margaret tracked productivity carefully through the transition:

Task	Before (avg)	After (avg)	Change
Receipt processing per client	45 min	12 min	-73%
Document review per return	30 min	18 min	-40%
Standard correspondence	15 min	4 min	-73%

“The time savings are real,” she said. “But the bigger benefit is staff morale. Nobody wants to type receipt data all day. Now they do work that actually requires thinking.”

What Didn’t Work

Not everything succeeded.

An AI tax research tool promised to answer technical questions. Margaret found its answers were often outdated or incomplete.

“It would confidently cite legislation that had been amended. Or miss recent ATO rulings. We stopped using it after a few close calls.”

An automated client communication bot was abandoned within a week. “Clients want to talk to humans, especially about tax. The bot felt impersonal and couldn’t handle anything outside its scripts.”

The Learning Curve

Implementation wasn’t instant. Margaret estimates six weeks before the team was comfortable with new processes.

“There’s a period where it’s actually slower because you’re learning the tools while doing the work. We had to push through that.”

Training came from vendor documentation, YouTube tutorials, and trial and error. No formal courses existed for the specific tools they chose.

Cost

Total investment for the transition:

• Software subscriptions: $340/month (ongoing)
• One-time setup and data migration: $2,400
• Lost productivity during learning curve: ~$3,000 (estimate)

“We made it back in the first tax season,” Margaret said. “Processing the same number of clients with the same team, but finishing earlier and with fewer late nights.”

Advice for Other Local Practices

Margaret’s recommendations for small professional practices considering similar changes:

Start with one process. Don’t try to transform everything at once. Pick the most painful task and fix that first.

Budget for mistakes. Early AI outputs need checking. Build extra review time into the first few months.

Keep clients informed. “I told my regulars we were trying new tools. Most were supportive. A few older clients needed reassurance that humans were still doing the important work.”

Don’t believe marketing claims. “Every tool promises to revolutionise your business. Most do one or two things well. Figure out what you actually need before buying.”

What’s Next

Margaret is cautiously exploring more advanced applications—AI that can identify potential deductions clients may have missed, or flag unusual patterns that warrant investigation.

“But I’m not rushing. The fundamentals of accounting don’t change because of AI. Understanding your clients’ situations, applying professional judgment, staying current with regulations—that’s still the job. The AI just handles some of the grunt work.”

Local Support

For St George and Sutherland Shire businesses considering similar technology adoption, several resources exist:

• Hurstville Library tech help desk — Free basic technology support
• Business Connect NSW — Subsidised advisory services
• Local chambers of commerce — Networking with businesses who’ve made similar transitions

Margaret’s practice, Liu & Associates Accounting, is located on Forest Road in Hurstville and has been serving the local community since 2004.