← all posts
§021 · BlogFinance

How AI Agents Can Reduce Month End Closing Time.

Close is slow because it is a chain of waiting, not a chain of thinking. Agents collapse the waiting and leave the judgement where it belongs.

7 min read

Days to close
3-4 days
from 8-12 days
Manual journal entries
Under 40
from 200+
Reconciliation
4 hrs
from 3 days

Ask a controller why close takes ten days and they will not say the work is hard.

They will say they are waiting.

Waiting on bank feeds, on expense reports, on one accrual from one manager who is on leave.

Close is not a computation problem. It is a coordination problem.

Agents are unusually good at coordination.

§01 · The problem

Where the time actually goes.

  1. 01Chasing accruals from department owners
  2. 02Matching transactions line by line
  3. 03Recurring journal entries keyed by hand
  4. 04Intercompany reconciliation across entities
  5. 05Variance explanations written from scratch
  6. 06Version control on the close checklist

Almost none of the close calendar is analysis. Most of it is collection, matching, and follow-up.

§02 · The mechanism

The AI workflow.

  1. 01

    Period end triggers the close

    Trigger
  2. 02

    Agent requests accruals from owners

    Automation
  3. 03

    Transactions matched automatically

    AI Agent
  4. 04

    Only exceptions raised to the team

    AI Agent
  5. 05

    Variance commentary drafted

    Generative AI
  6. 06

    Controller reviews and signs off

    Human sign-off

A person closes the loop: the workflow ends on a human or a check against a defined outcome.

Trigger
The event that starts the workflow
Automation
Deterministic. No model involved
AI Agent
Model extracts, classifies, matches, or routes
Generative AI
Model drafts something a human will edit
Human sign-off
Final approval before anything ships

§03 · The difference

Before vs after.

BeforeAfter
Email every owner for accrualsRequests go out automatically on day one
Follow up for four daysAgent follows up until it has answers
Match transactions in a spreadsheetMatching runs overnight, unattended
Investigate every mismatch by handOnly genuine exceptions reach a human
Write variance notes from memoryCommentary drafted from actual movement
Close lands somewhere in week twoClose lands in the first week

§04 · The measure

Time saved.

TaskBeforeWith AI
Accrual collection4 days1 day
Transaction matching3 days4 hrs
Variance commentary1 day1 hr
Total close cycle8-12 days3-4 days

Continue reading

Tell us what you’re building.

Share a few details about what you need, and we’ll get back to you with a quotation.