Your Agents
Six deterministic agents watching the close. They propose — nothing changes until a human accepts.
Publication gate
- Group COA resolution104 accounts checked, 0 unresolved
Every reported trial-balance account resolves to the group chart
- FX consistency5 contributions checked, 0 off
Translated amounts reproduce from local amount × exchange rate
- Elimination integrity8 journals checked, 0 invalid
Every elimination posts to group accounts, on two distinct legs
- Variance arithmetic21 rows checked, 0 off
Movement columns recompute: variance = current − prior quarter
- IC matching8 pairs, 1 unmatched, 0 unexplained
Every intercompany pair is matched, or its difference is explained
- Mapping governance18 confirmed, 7 in review, 3 unmapped of 28
All local accounts confirmed onto the group chart, none dangling
Agent roster
Reads every local-to-group mapping and flags accounts with no group home, proposals still waiting on a human, and machine suggestions whose confidence is too low to trust blindly.
Reads every intercompany pair, splits clean matches from disagreements, and sizes each gap. Anything unmatched entirely gets its own alarm — there is nothing to net against.
Checks each elimination posts to real group accounts on two distinct legs, then cross-reads the IC matcher's findings for intercompany positions no journal covers.
Recomputes each audit-trail contribution from local amount and rate, confirms the translation reserve arithmetic, and sizes the currency exposures that move consolidated figures most.
Counts who is in the close and on what basis — full consolidation, equity method, NCI — and flags structural facts the consolidation logic must respect, like equity-method investments and partial trial-balance extracts.
Reads the six deterministic publication checks plus what the other agents found, and issues the only verdict that matters: publishable or held. When held, it drafts the sign-off request listing exactly what blocks release.
Findings (14)
3 local accounts have no group mapping at all
These accounts sit outside the group chart, so their balances cannot flow into any consolidated line. Every publication figure that should include them is understated until someone maps them.
- Accounts
- SUB-CN 7890 (其他), SUB-SG 9990 (Suspense), SUB-JP 9999 (仮払金)
- Share of mappings
- 10.7%
IC001: HQ-US and SUB-CN disagree by $45,000 on Management Fee
Both sides booked the same management fee in the same currency, but for different amounts. One side is wrong, and the elimination can only be as good as the number it is built on.
- HQ-US booked
- USD 3,445,000
- SUB-CN booked
- USD 3,400,000
- Difference
- $45,000
- Mismatch type
- amount
IC008: License Fee of $450,000 from HQ-US has no counterparty entry
HQ-US booked a license fee to SUB-JP, but SUB-JP booked nothing on the other side. A one-sided intercompany position cannot be eliminated — it will distort group revenue until the other side posts or the charge is removed.
- Amount
- $450,000
- From → To
- HQ-US → SUB-JP
- Status
- Unmatched — no counterparty posting
Publication held — 2 of 6 checks failing
The close cannot be published while deterministic checks fail. These are arithmetic and governance gates, not judgement calls: fix the mappings and the unmatched IC position, and the gate releases on its own.
- Failing checks
- IC matching; Mapping governance
- Blocking findings upstream
- 3
- Checks passing
- 4 of 6
The suggestion: Q1 2026 publication sign-off (Draft) — accepting records the request, nothing more. Agents propose; humans sign.
7 mapping proposals are waiting on a human
The mapper proposed group accounts for these, but nothing is confirmed. Three of them have no proposal at all — those need an accountant, not an approval click.
- Awaiting review
- SUB-SG:4100, SUB-CN:6500, SUB-CN:6600, SUB-JP:4000, SUB-JP:1100, SUB-JP:5000, SUB-DE:5200
- No proposal yet
- SUB-CN:6600, SUB-JP:5000, SUB-DE:5200
2 smaller IC differences totalling $2,900
Individually immaterial, but they point at process noise — timing or rounding between entities. Cheap to clear now, annoying to explain in an audit later.
- Pairs
- IC003 ($2,100), IC007 ($800)
- Combined gap
- $2,900
2 blocking IC issues have no clean elimination path
The elimination layer assumes both sides of an IC pair agree. Where they do not — or one side is missing entirely — no journal can be drafted honestly until the underlying postings are fixed.
- Open IC issues
- IC001, IC008
- Eliminations on file
- 8
JV-IN (India Joint Venture) is equity-method at 45% — keep it out of line-by-line consolidation
An equity-method investment contributes one line (share of profit and net assets), not a full trial balance. If its balances are folded in line-by-line, revenue and assets are both overstated.
- Ownership
- 45%
- Method
- Equity
- Currency
- INR
All 8 trial-balance extracts are partial — P&L close lines are not in the feed
Every entity's extract shows credits above debits by exactly the profit lines that have not been closed into equity. The consolidation engine treats these as balance-sheet-plus-revenue extracts, which is fine — but nobody should tie these files back to a signed-off TB without the close entries.
- Largest gap
- JV-IN: INR 41,400,000
- Entities affected
- 8
10 mappings scored below 90% confidence
Low confidence is not an error, but it is where wrong mappings hide. Worth a skim before the close is called done.
- Weakest
- SUB-CN:7890 (0%), SUB-SG:9990 (0%), SUB-JP:9999 (0%)
All 8 elimination journals are well-formed — $11,337,500 eliminated
Every journal posts to group chart accounts, on two distinct legs, with a positive amount. The elimination layer itself is structurally sound.
- Journals
- 8
- Total eliminated
- $11,337,500
- Journal types
- 4
SUB-CN carries the largest translation effect in the audited account
Its CNY contribution compresses the most on translation — a reminder that JPY, CNY and INR exposures, not operating moves, drive much of the period-on-period movement in translated balances.
- Local amount
- CNY 10,800,000
- Translated
- $1,500,000
- Rate
- 0.1389
Every audited translation reproduces within rounding (5 checked)
Each entity contribution in the audit trail equals its local amount times the stated rate, within 1% rounding. The FX math is reproducible end to end.
- Account audited
- 1100 Trade Receivables
- Contributions
- 5
- Off beyond rounding
- 0
8 entities in scope: 1 parent, 6 fully consolidated, 1 equity-method
Entity coverage and ownership are the skeleton the whole close hangs on — eliminations, NCI and translation all key off this list.
- Fully consolidated
- SUB-SG, SUB-HK, SUB-CN, SUB-JP, SUB-DE, SUB-UK
- Equity method
- JV-IN (45%)
- With NCI
- SUB-CN (20%), SUB-UK (30%), JV-IN (55%)