What teams build in Prowork for accounting and close work

Accounting work tends to get automated early, largely because the rules already exist somewhere. The team knows what the tolerance is, how the coding should go and who owns each kind of exception, so what is missing is rarely the logic and almost always a way to apply it to real volume.

Of the 139 companies whose usage went into this study, 82 run some version of this work in Prowork, spread across 1,559 flows that were active in the last 90 days. Grouping those flows by what each one actually does gives the breakdown below.

Accrual Calculations is the largest of those patterns at 415 flows across 38 companies, with Cash Reconciliation accounting for another 337 across 43 companies. The grouping follows what a flow does rather than which team happens to own it, which is why one company often turns up in several of these rows at once, and the businesses building the most of them work in apparel, health and beauty, and food and beverage.

Three of those flows in more detail, described by what each one does rather than by who built it. Company names are withheld, so each is identified only by the size and category of the business running it.

A $100M+ software company. Captures bank ACH confirmation details and standardizes the key totals needed for reconciliation. Helps finance verify cash movement files faster, confirm expected totals and item counts, and reduce manual review of bank notifications. Built on API, email attachments.

A $100M+ operations-heavy business. Standardizes trial balance data from multiple property management systems into a consistent accounting import format. Applies property and GL account mappings so the accounting team can generate upload-ready journal files with less manual cleanup and fewer posting errors. Built on SharePoint, email attachments.

A mid-market operations-heavy business. Calculates monthly loyalty points liability by consolidating loyalty activity, liability balances, redemption payments, and historical redemption assumptions across multiple storefronts. Built on email attachments.

Read enough of these flows and the same shape keeps appearing. A flow pulls the same records from each system that holds a version of them, standardizes whatever identifiers are needed before those records can be joined at all, applies the comparison or the calculation as explicit logic, and then labels every row with an outcome so that the handful needing attention can be routed to whoever is able to act on them.

What differs from one flow to the next is which single step genuinely calls for interpretation. Reading a supplier PDF, resolving a merchant name that never quite matches the ledger, or deciding which category a vague line description belongs in are all handled by AI steps inside the flow, while the matching rules, the tolerances and the thresholds around them stay explicit, since those are the parts a controller or an auditor will eventually want to read for themselves.

Most of these run on a schedule or fire from a trigger rather than waiting for someone to remember to open them, and that is largely what separates a report describing what happened last month from a process that surfaces the problem while there is still time to do something about it.

The sources they read from are worth noting, because they are rarely the tidy ones. Across the documented flows in this category the most common inputs look like this:

Email attachments and spreadsheets sit at the top of that list on most pages, which is a fair description of where operational data actually lives once it leaves a system of record.