I Set Up the Books Before There Was Any Revenue. That Was the Point.

Clarent has zero revenue. It has a bookkeeping system anyway, and building it now — while there's nothing to book — is the only reason it'll be trustworthy in April.

I've watched the alternative up close. Someone starts a business, pays for things on whatever card is handy, and a year later hands a shoebox to an accountant who charges by the hour to reconstruct intent. I wrote about going deeper on my own taxes in the spring. This is that idea applied from day one.

What I didn't buy

QuickBooks. Xero. A bookkeeper. A numbered general ledger with 10000 Cash and 60000 Expenses.

A single-member LLC is disregarded for federal tax. Everything lands on Schedule C of my personal return. A five-digit chart of accounts earns its keep only when you have accrual accounting, a bookkeeper, or an S-corp election. I have none of those. So the chart of accounts is the Schedule C line list itself: Advertising (line 8), Legal & Professional (17), Office (18), Utilities (25), Software & Subscriptions (27a), and so on.

Pick the chart of accounts your tax return already uses. Every category maps to a line. Nothing to translate in April.

The system

• One spreadsheet, seven tabs. Summary, Categories, Expenses, Equity, Home Office, Missing Receipt Log, and a Guide tab that explains the rules to future me. It lives in the Finances shared drive.
• Receipts named so they sort themselves. YYYY-MM-DD_Vendor_Amount_description.pdf. Open the folder and it's already chronological. No renaming, no searching.
• Category and tax treatment are separate columns. A formation fee is coded to "Taxes, Licenses & Filing Fees" and flagged as an organizational cost. The category never has to carry the tax rule.
• An Equity tab. Schedule C has no equity section, so the ledger has a small one. Every dollar I spent personally before the bank account existed is recorded as a capital contribution, not an expense that vanished. The running member capital balance is what proves the entity is real.
• Business-use percentages, written down once. Phone, internet, the AI subscriptions. I picked the number that's true, wrote a one-paragraph memo on why while the reasoning was fresh, and left it alone. A percentage that drifts upward after you see how small the deduction was is the thing that draws attention — not the percentage itself.

The mistake I already made

Within a week I had three copies of the ledger: a Google Sheet, an Excel file in iCloud, and the shared-drive copy. They drifted. Rows went missing from one and not the others.

Maintaining parallel ledgers by hand is how ledgers go wrong, and it happened to me in under ten days. Now there's one file, in one place, and the others are labeled ARCHIVE. If you take nothing else from this post: one system of record, decided on day one.

Where Claude fits

Two places. First, the design — I described what the business looked like and what I'd need at tax time, and we iterated on the structure until it was simple enough that I'd actually maintain it. Second, the monthly loop: Mercury's read-only connector lets Claude pull the bank feed and reconcile it against the ledger, flag anything unmatched, and tell me what's missing a receipt. The bank feed is the truth; the ledger is the explanation. Claude checks that they agree.

Total tooling cost: zero dollars a month. I'll revisit when there's real revenue, a second member, or a CPA who wants a live file.

The part I'm handing to a professional

Everything above is bookkeeping, not tax advice. The startup-cost treatment, the business-use percentages, the home office, and the question of when the business officially "began" all go to my CPA before the first return is filed. The system's job is to make that conversation short.

If you're a solo operator and your books are a shoebox, ask me anything — I'll tell you exactly what I'd set up first. And if you run a small business and think I've overcomplicated this, I want to hear that too.