Vendor bills
The single largest block in most companies. Read, coded, matched, and routed — typically the first workflow customers automate and the one with the clearest return.
AI capability
Data entry is the largest block of low-value work in a finance function and it scales linearly with growth, which almost nothing else in finance does. A realistic target is removing eighty to ninety percent of it — and being specific about the remainder is more useful than claiming all of it.
What it does
The single largest block in most companies. Read, coded, matched, and routed — typically the first workflow customers automate and the one with the clearest return.
Card transactions matched to photographed receipts, coded, and policy-checked, so employees stop assembling reports and finance stops chasing them.
Feed transactions matched to open items daily rather than keyed and reconciled monthly.
Drafted from calendar and activity for confirmation rather than reconstruction, which is where time accuracy actually improves in service businesses.
Captured once at signature and feeding billing and revenue recognition, rather than re-entered by finance from a PDF a quarter later.
Mapped to accounts and dimensions automatically instead of arriving as one lump entry somebody splits by hand each month.
When we measure a finance function before an engagement, the distribution is consistent enough to be worth stating. Roughly half of manual entry is vendor bills. Another fifth is expenses and receipts. Bank and card reconciliation is around fifteen percent, timesheets and payroll mapping around ten, and everything else — journal entries, contract data, customer setup — makes up the remainder.
That distribution is why we recommend starting with accounts payable regardless of industry. It is the largest single block, the most repetitive, and the one where straight through rates climb fastest because the same forty vendors produce most of the volume.
Eighty to ninety percent of keying is removable for a typical mid-market company within two quarters. The remaining ten to twenty percent is genuinely resistant, and it is worth understanding why rather than treating it as a gap to close later.
Some of it is the long tail — vendors you use once, documents in unusual formats, novel transaction types. Some of it is judgement wearing the costume of data entry: the coding decision that is actually a policy decision, the allocation that requires knowing something not written on the document. And some of it is deliberate friction we would not remove even if we could, because a person looking at it is the control.
The pitch is usually headcount reduction and that is not what we observe. Most customers keep the same finance team and change what it does: closing faster, analysing variances that were previously never investigated, and doing the work that was permanently deferred because keying consumed the week.
The measurable outcomes tend to be close duration falling, DSO improving because collections finally get attention, and the finance team stopping being a constraint on growth. Whether you convert that into headcount savings is your decision rather than an automatic consequence.
Limits
Every capability page on this site carries one of these, because a feature described without its boundaries is a claim rather than a description.
Deciding whether a cost is capitalised, whether an obligation is distinct, or whether a variance is acceptable is work that looks like keying and is not. It stays.
One-off vendors, unusual documents, and genuinely novel transactions will keep reaching a person. That is roughly ten to twenty percent of volume and it is the design rather than a shortfall.
A supplier who emails a photograph of a printed spreadsheet is a supplier relationship problem, not an extraction problem. We surface it rather than absorbing it.
Questions
Tell us your monthly volumes by document type and we will estimate what is removable and what is not.