Why an accounting firm built its own Financial OS

I run an accounting firm. For years we did what every firm does: we assembled a stack. QuickBooks Online for the ledger. A practice-management portal for close checklists and client requests. A document-collection tool for receipts and statements. An e-signature subscription. Email threads to hold it all together.

Every one of those tools billed us separately, most of them per client, and none of them talked to each other properly. When QuickBooks raised prices again, we did the math on what the stack actually cost us per client per month — and then we did something unreasonable. We built our own platform.

This article is the honest version of why, because "we built our own software" usually deserves skepticism.

The problem was never one tool

QuickBooks Online is a competent ledger. Most of the tools in a firm stack are competent at their one job. The problem is the seams:

• The ledger doesn't know a close checklist exists, so month-end lives in a separate tool or a spreadsheet.
• The document tool collects a receipt, but attaching it to the transaction it supports is somebody's job.
• The portal asks the client for a bank statement; whether it arrived is a status you check, not a fact the books know.
• The e-sign tool has its own client list, which drifts out of sync with the other three client lists.
• Every seam is a login, a subscription, and a place where work silently falls on a human.

At a handful of clients you don't feel it. At a few hundred, the seams are most of the job. Our staff spent real hours every week moving information between products that each claimed to be saving us time.

The moment the math flipped

Two things pushed us from complaining to building. The first was pricing: per-client software costs kept ratcheting up across the whole stack at once, and multi-entity clients — a landlord with six LLCs, say — were charged as six subscriptions for what is really one set of books to us.

The second was watching tools we depended on get retired or degraded underneath us. When a product you built your workflow around can disappear on a vendor's schedule, "we don't control our own stack" stops being an abstraction.

We had the unusual advantage of being our own first customer. A firm knows exactly where the hours go. So we specified the platform we always wished existed, and built it.

What Woven actually is

Woven is a Financial Operating System: one client-centric platform where the ledger and the work around the ledger are the same product.

• A real double-entry general ledger — journal entries, sub-ledgers, audit trail. Not a layer on top of someone else's books; the book of record itself.
• Bank feeds and statement import built in, with reconciliation that checks itself against the statement's own printed subtotals.
• Document intake inside the books — receipts arrive by portal, upload, or a per-client email address, and land attached to transactions.
• Month-end close, client requests, and firm tasks as first-class features of the ledger, not a separate practice tool.
• A client portal, e-signature, reporting packages, and multi-entity consolidation — the whole stack, one login, one client list.

On AI, our position is deliberately unfashionable: AI in Woven is assistive and human-verified, never autonomous. It proposes categorizations, reads documents, drafts responses — and a person approves the work, because these are somebody's books and possibly somebody's tax return. Client data can be excluded from AI features entirely with a per-client switch, and personally identifiable information is redacted locally before any document goes to a model.

What we deliberately didn't build

We did not build an "AI accountant." We do not believe the technology is there for unsupervised bookkeeping, and we run a firm — we know what unreviewed books cost to fix. We also didn't build a thin wrapper over QuickBooks: if the ledger isn't yours, every one of the seams above eventually comes back.

Where this stands today

Woven runs real books today — including our own firm's clients, which is the most honest QA process we know: our bookkeepers hit the rough edges before yours do. We're onboarding beta firms now, and migration imports your chart of accounts, customers, vendors, balances, and history from QuickBooks or Xero, so switching means verifying, not re-keying.

If you run a firm and any of this sounds familiar, we'd genuinely like to show you. We host a live walkthrough for the beta cohort — no slides, the actual product on seeded books — and you can register for the next session or [book a demo](/demo) directly.

And if you're happy with your stack: keep it. This exists for the firms doing the math we did.