Cassia

Small businesses and their bookkeepers spend hours every week on bank reconciliation, the tedious process of matching bank transactions to contacts, accounts, and invoices. It's repetitive, error-prone, and eats into time that could be spent actually running the business.

For bookkeepers and accountants managing multiple clients, the problem compounds. Tax season becomes a crunch of catching up on months of unreconciled transactions, and the manual nature of the work means margins stay thin no matter how many clients you take on.

Cassia was an AI bookkeeping assistant that logged into a business's accounting software and learned their unique bookkeeping patterns, building a custom AI model for each business.

When new bank transactions came in each day, Cassia used that model to automatically match transactions to the right contact and account (like "Sales" or "Electricity") or to existing invoices and bills. Rather than acting blindly, Cassia gave users a 12-48 hour review window before reconciling, letting them review, adjust, or cancel any match.

Once the review window closed, Cassia logged back into Xero and completed the bank matching automatically. The result was up to 90% of bank reconciliation handled without manual input.

We ran Cassia with local businesses and bookkeepers, proving the model worked in production. The product was live, matching real transactions, and delivering on the promise of automated reconciliation.

Then Xero shipped their own AI-powered bank reconciliation features natively within the platform. When the accounting software itself builds the feature you're selling on top of it, the strategic calculus changes. We made the call to sunset Cassia and pivot our focus.

The validation was real though: we'd identified a genuine pain point, built a working product around it, and got it into the hands of real users before the incumbent moved.