CloseCheck — shorten review of monthly-close spreadsheet exceptions

The problem

Finance teams manually inspect recurring exports for missing fields, inconsistent labels, and unusual values before monthly close.

The idea

Why now

Docker Agent can package a configurable agent workflow; Laya can return typed decisions and scores rather than prose (docker-agent-ai-agent-builder-and-runtime-by-docker-engineering, laya-open-source-decision-engine).

What it combines

Docker Agent packages the review workflow, while Laya classifies candidate exceptions. A separate ordinary web table makes each result inspectable without requiring ChatGPT plugin registration.

How to build it

1. Select a recurring export. Agree on one CSV template, exception definitions, and the current review-time baseline. (1 day)
2. Preserve evidence. Parse rows and retain column names, values, and source row numbers. (1 day)
3. Classify candidates. Run defined checks and Laya judgments through a Docker Agent workflow. (1 day)
4. Pilot review. Build a web table for dispositions and compare results with the controller’s existing review. (1 day)

Who benefits

Controllers

Reach an evidence-linked exception list sooner.

Finance analysts

Spend less time sorting candidate exceptions.

How it makes money

Model

Company subscription

Pricing

$299 per month for one close workflow

Wedge

One controller’s recurring CSV export

Channels

Direct outreach to controllers · Controller referrals within finance teams

Distribution

A mid-market controller pays. Start with one recurring CSV export, reached through direct outreach to controllers who own the close process.

Why it wins

CloseCheck presents the triggering cell, column, rule, and source row alongside a recommendation. It neither edits the ledger nor substitutes for controller sign-off.

Risks

A missed exception could matter more than saved review time. Run alongside the existing review, measure misses against controller findings, and retain its normal sign-off.

Strongest objection left standing

Even with visible cell evidence, controllers may not trust model judgments enough to shorten their established review.

Build it with

• Docker Agent: AI agent builder and runtime by Docker EngineeringPackages and runs the configurable review workflow.
• Laya: open-source non-autoregressive decision engine (Apache 2.0)Supplies typed judgments and scores for candidate exceptions.

Repo to start from

closecheck - evidence-linked close exception review

Evidence

• Docker Agent
• Laya

Debated by openai/gpt-6-sol, anthropic/claude-sonnet-5.5, xiaomi/mimo-v2.6-pro, qwen/qwen3.8-max-0902, moonshotai/kimi-k2.5. Judged by JEV.