Audit Agent
Collects, labels, and packages audit evidence on request.
~900 hours1 (modeled outcome) saved per audit year
Finance agent
GAThe Close Agent is a governed AI agent that runs the month-end checklist. Chases what is late. Records who signed.
How it works
Sequences close tasks by dependency, prepares reconciliations and recurring accruals, nudges owners, routes sign-offs, and keeps a live status the controller can trust.
Load your task list with owners, reviewers, and dependencies. The agent connects to the GL, subledgers, bank feeds, and intercompany, and learns prior-period timings.
Bank and subledger reconciliations that tie are prepared with support. Recurring accruals are drafted from prior periods and open POs. Prepaid and fixed-asset roll-forwards are staged.
Owners get reminders before due dates and nudges after. Escalations go to the controller at the threshold you set. Blockers upstream are visible to everyone downstream.
Sign-offs route to named reviewers and are timestamped. A daily status report replaces the tracker. A post-close retrospective shows where the days went.
| Day | Task | Owner | Status |
|---|---|---|---|
| Day 1 | Sub-ledger cutoff · AP, AR, payroll | M. Reyes | Complete |
| Day 1 | Bank reconciliations · 14 accounts | Agent | Complete |
| Day 2 | Accrual review · 22 recurring | Agent | Complete |
| Day 2 | Intercompany elimination | S. Adeyemi | In progress |
| Day 3 | Flux analysis + commentary | Agent | Awaiting sign-off |
The agent contract
Every Meridian agent runs under the same six terms, enforced at the Gateway below the agent. Here is what each one means for the Close Agent.
Scoped to one workflow
Runs the close checklist and prepares mechanical work. It never posts a journal entry.
Reads only permitted data
Reads the GL, subledgers, bank feeds, and intercompany balances in the entities you assign.
Every action logged
The close log is immutable, timestamped, and retained as audit evidence.
Consequential actions require human approval
A preparer posts and a reviewer approves. Unexplained reconciliation differences never auto-close.
Measurable outcome
Days to close, late tasks, and reconciliation prep hours reported per entity.
Runs on your chosen model
Model choice per workspace. Preparer and reviewer separation is enforced below the agent.
Scope
What the Close Agent reads, what it does on its own, and where a person signs.
Outcomes
Modeled outcomes from design-partner deployments. The Registry reports your actuals from the first week an agent is live.
shorter month-end close1
manual reconciliation preparation1
of close tasks with owner, evidence, and timestamped sign-off
close status instead of a Friday spreadsheet
“The auditors asked for the lineage record on the first package and never asked again. We went through interim fieldwork without pulling anyone off the close.”
shorter close by the second cycle at Castellan Financial1
In production at
Integrations
Connected through prebuilt connectors and zero-copy warehouse access in Data Fabric. The agent acts as the user and inherits each system’s permissions on every read.
Credits
Included on every plan: Starter with 5,000 credits per month, Growth with 30,000, and Enterprise with custom pools. Credits are consumed per completed action, never per message. Usage beyond the plan is billed at $0.12 per credit.
| Action | Credits |
|---|---|
| Close task orchestrated | 1 |
| Reconciliation prepared with documentation | 2 |
| Accrual drafted | 2 |
| Close package with status and retrospective, per entity | 10 |
Questions
Answered the way buyers ask them. Each answer is also published as structured data.
Mostly by removing waiting. The Close Agent knows the dependency graph, prepares the mechanical reconciliations and accruals before anyone asks, and chases late tasks so downstream owners are not idle. Design partners took about 3 days out of a 10-day close in the first two cycles.
No. It drafts entries and reconciliations. A preparer reviews and posts. A reviewer approves. The separation your auditors expect is preserved, and the agent's drafts are labeled as such in the log.
Yes. It can run inside a close management platform through its API, or it can be the checklist for teams still on spreadsheets. Either way the task list, owners, and dependencies are the source of truth.
They are flagged with the difference, the likely causes the agent can see, and the transactions involved. A human investigates and resolves. The agent never closes a reconciliation with an unexplained difference.
Related agents
Agents most often deployed alongside the Close Agent, on the same data and the same approval policy.
Get started
Start with one workflow, one approver, and one number to move. Most design partners were live in five weeks.