Agent Library · By function · Receivables · Close · Controls
Month-end assembles itself. Your controllers control.
Finance is the function where automation has to be auditable or it is worthless. Every action these agents take leaves a trail an auditor can follow, and the three decisions that move money never happen without a human.
Receivables slip quietly
Nobody owns the third reminder. Days sales outstanding stretches, and the cash was always yours.
Scope creep goes unbilled
Work happens in scattered tools and chats. At month-end nobody can prove what was out of scope, so it isn't invoiced.
Close is a five-day build
Your most senior finance people spend the first week of every month compiling instead of controlling.
The agents
6 agents. Each one earns its autonomy.
Agent
What it does
Autonomy
Human gate
AR Chaser
Drafts and schedules invoice-reminder sequences, escalating in tone and channel as they age.
Notify
Tone reviewed for marquee clients
Three-Way Match
POs, GRNs and vendor invoices extracted, matched, and queued for approval — with the mismatch explained.
Review
Any payment approval stays human
Scope Sentinel
Watches delivered work against contracted scope and flags billable out-of-scope work while it is still provable.
Flag only
—
Close Assembler
Month-end packs built from source systems automatically. Your team reviews the numbers, not the formulas.
Review
—
Exception Surfacer
Pricing errors, stock mismatches, duplicate payments and blocked entries found and reported the hour they occur.
Flag only
—
Reconciliation Checker
ERP, CRM and books compared continuously; discrepancies flagged before they compound into a restatement.
Auto
—
Autonomy key — Auto: executes above 95% confidence. Notify: executes at 85–95% with notification and undo. Review: always queued for a human. Flag only: never acts, alerts a person.
What changes
Three shifts, not a feature list.
The cash you already earned arrives sooner
Nothing about the invoice changes. The follow-through stops depending on someone remembering.
Unbilled work becomes billed work
Scope creep caught in the week it happens is revenue. Caught at month-end it is an argument.
Every automated step leaves a trail
Logged, attributable, reversible — automation an auditor can sign off on.
The eval bench
What these agents have to pass before they touch real work.
- Match accuracy on 100 historic three-way match cases, including the ones that legitimately failed
- Zero invented figures: every number in an output traces to a source document, or the agent fails and flags
- Reminder tone graded against your own past correspondence, per client tier
Written against your data, not ours. The bench is built during Blueprint and runs on every deploy from then on.
Related
Where this connects.
See it on your Tally.
A Blueprint maps which of these agents pays off first in your exact setup — and which to leave alone.
Apply for a Blueprint