Automation for the work that repeats

What this actually is

Not a product, and not a platform you log into. It is custom work built against a specific business's chart of accounts, vendors, and workflow — because that is the part generic software cannot do. Off-the-shelf tools make you reshape your process to fit them. Building on top of your own system means the process stays as it is and the repetitive parts of it stop needing a person.

It sits on top of an existing bookkeeping relationship. Someone has to know what the books are supposed to look like before it is worth automating anything that touches them, which is why this is offered to clients rather than as a standalone service.

What can be built

• Document intake. Invoices, receipts, and statements pulled from email or a shared drive, read, and filed against the transaction they belong to — instead of a monthly hunt through an inbox.
• Recurring coding. Vendors and transaction patterns that always code the same way, applied consistently, with anything unfamiliar held back for a person to look at.
• Exception flagging. Transactions that do not match any expected pattern surfaced early — a duplicate payment, a charge that doubled, a vendor that has never appeared before.
• Cross-system reconciliation. One system's records checked against another's on a schedule: a payroll provider's reports against the ledger, a back-office system against the accounting file, a rent roll against deposits.
• Reporting. The same monthly package assembled from the same accounts every time, delivered on a fixed date rather than when someone gets to it.
• Internal tools. Small applications for the things a business does constantly and awkwardly — a form that produces a correctly coded entry, a lookup that answers a question staff currently ask the bookkeeper.

What does not get automated

This matters more than the list above, because the failure mode of automation is not being wrong occasionally — it is being confidently wrong at scale, in a system of record, without anyone noticing for months.

• Judgment calls. Whether an ambiguous expense belongs in one account or another, whether a cost is a repair or a capital improvement, how an unusual transaction should be treated.
• Anything carrying professional responsibility. Trust and escrow reconciliations get signed off by a person. So does the monthly close.
• Anything a regulator or a CPA will rely on without a person having checked it first.

Every automation is built so a person reviews what it did before it becomes the record. The books stay the bookkeeper's responsibility, not the software's — automation changes how long the work takes, not who is answerable for it.

How QuickBooks integration works

QuickBooks Online publishes an API that lets an authorized application read and write transactions, bills, invoices, vendors, and reports. Access is granted by the account owner, scoped to what the tool actually needs, and can be revoked at any time from inside QuickBooks. Nothing is installed on your machines and no passwords are shared — the connection is authorized the same way any approved QuickBooks app is.

Where a business runs something other than QuickBooks — a brokerage back-office platform, a point of sale, a club or property management system — what is possible depends on what that system exposes. That is the first thing to establish, and it is a short conversation.

How an engagement runs

1. Start from a bottleneck, not a technology. Something specific that takes hours every month and shouldn't.
2. Scope and fixed price. Quoted as a project, separate from the monthly bookkeeping retainer, so it is a decision you make once rather than an open meter.
3. Built, then run alongside the manual process until it demonstrably agrees with it.
4. Handover. You own what is built and it keeps working if the engagement ends. No hostage software.

[ ] Data handling — required before this page goes live. This section must state plainly what happens to client financial data: where it is processed, whether any of it reaches a third-party AI provider, whether it is retained, and an explicit commitment that it is never used to train anyone's model. These are promises about other people's money, so they have to be decided by the practice rather than drafted here. A law firm's trust data or a brokerage's escrow will not be handed over without them.

[ ] Who builds this. Name, background, and how the technical side of the practice is presented — deferred until the new About page introduces the second person. Until then this page describes a capability with nobody attached to it, which is the weakest part of it.