EP549: Sam Leon - AI Does The Heavy Lifting: How One CPA Automated His Whole Workflow

See what the team at The Successful Bookkeeper has on right now →

If you have ever wondered what it actually looks like to build AI into your accounting practice — not in theory, but day-to-day — this episode is for you. Sam Leon, CPA and founder of TaxWeave, joined guest host Louie Prosperity to share exactly how he went from experimenting with ChatGPT to building his own secure AI platform for tax professionals. The conversation is practical, honest about the risks, and full of ideas you can act on right away.

Chapters

• [00:00] Introduction and Sam's Background
• [02:45] Why Sam Chose Claude Over Others
• [06:30] Claude Code: Beyond Just Chatting
• [09:30] Workflow Automation for Bookkeepers
• [12:30] The Data Security Problem
• [16:30] Building TaxWeave as a Secure Platform
• [20:00] Top AI Tasks to Tackle First
• [24:00] Keeping Up as AI Models Evolve
• [26:30] How Sam Checks His AI Outputs
• [28:30] Where to Find Sam and TaxWeave

Here’s a special gift for YOU. Visit Plooto.com/TSB and try it for FREE for two months! That’s a FREE 30-day trial, plus an extra free month for Successful Bookkeeper podcast listeners!

Don’t miss out, sign up TODAY!

From Curiosity to Real Workflows

Sam has been watching technology waves roll through the accounting profession for years — cryptocurrency, machine learning, practice management software. When AI chatbots arrived, he started testing them immediately, but it was a specific feature of Claude that changed how he worked: the ability to connect the tool directly to files on his computer. "That's how I got started with a lot of the more time-saving workflows that I was doing," he explains. The shift from chatting with AI to feeding it real documents opened the door to genuine time savings rather than novelty.

The Tasks Best Suited for Automation

Sam is direct about where AI earns its keep: document summarization, transaction categorization, and building QuickBooks rules. He recently spoke with a bookkeeper who used AI to generate an extensive list of transaction rules that could be imported straight into QuickBooks — a task that would have taken hours to build manually. His broader advice is to start with tasks that are data-heavy and repeatable, then layer AI on top of the tools you already use rather than replacing them outright.

Handling Sensitive Data the Right Way

This is where Sam slows down and gets careful, and so should you. For tax professionals in the US, IRS Code Section 7216 governs what client information can be shared with third-party systems — and most AI platforms have not yet been through the kind of SOC 2 certification process that established tax software has. Sam's personal rule: never upload anything containing Social Security numbers, EINs, or last names. For bookkeeping work, he sees a slightly lower barrier since transaction lists are often less personally identifying, but the principle holds. "Being more cautious than not — if you're unsure about it, probably take another step to make sure that information really doesn't have any personal identifying information on it." Upgrading to a Teams or Business plan on these platforms also adds a layer of data protection over personal accounts.

Building TaxWeave: From Workaround to Platform

The manual redaction problem — using Adobe to black out client details before running documents through AI — was slow and error-prone. That frustration pushed Sam to start building TaxWeave, a platform designed to handle the redaction automatically inside a secure environment, then run AI workflows on the cleaned documents. He built the initial prototype using Claude Code, which lets you describe what you want in plain language and generates the underlying code. "The barrier where you had to understand how to do that — it's removing that," he says. TaxWeave currently focuses on tax practice project management and client organization, with AI workflow integration being the layer built on top.

Keeping Up as the Models Keep Changing

AI models update constantly, and Sam's advice is to treat your workflows the same way you treat your engagement letters — review them periodically and adjust. He recommends using the Projects feature in Claude to maintain context across sessions, and revisiting your saved instructions whenever a new model version drops. You can even ask the AI itself what needs updating. For quality control, Sam still does his own research to verify AI-generated answers and checks outputs manually, paying attention over time to where the tool is weakest. "You can start to see what it's weaker on — what kind of things does it always miss — and what you can start trusting it with."

Links Mentioned

• TaxWeave: taxweave.ai
• Sam Leon on LinkedIn: search Samuel Leon CPA
• Claude (Anthropic): claude.ai
• PureBookkeeping: purebookkeeping.com
• Pluto (sponsor): pluto.com/tsb

About the Guest

Sam Leon is a CPA with 14 years of experience in tax accounting, specializing in corporate tax, startup tax issues, R&D credits, and individual returns. After building his solo firm, The Millennial CPA, he was recognized in the 2026 Accounting Today Best Firms for Technology Awards. He is currently developing TaxWeave, an AI-powered platform built specifically for tax practices, and is active in conversations across the profession about safe, practical uses of AI in accounting.