Juno would tie every balance-sheet account and sub-ledger to its source each day, draft the explanation and the adjusting entry for each difference, certify low-risk accounts inside the policy, and run the close calendar so people handle the exceptions.
It is workday 2 of the September close, 08:15, and J. Harrison, finance lead, opens the reconciliation queue. Overnight the agent tied out 388 balance-sheet accounts across SAP S/4HANA and the Oracle Fusion division, certified 351 inside their tolerances under the rule, and drafted an explanation and a journal for 37. Each draft carries the difference, the documents it found and a confidence.
The goods-receipt clearing account at plant 1210 is the one she opens first: 38 open items over 60 days, worth $186,000. Twenty-nine are receipts from Delta Fasteners Ltd still awaiting invoices, six are price differences on material 7741-X, and three carry an unmatched document. She approves the drafted reclassification for the six price differences, and S/4HANA answers with document 4900021877.
One click sends the 29 supplier items to D. Okafor, buyer, with the purchase order lines attached, because the invoices belong to his desk. The three unmatched items wait for the Thursday review. The clock on the rest of the queue reads 2 days 6 hours, and the close calendar shows bank reconciliations done, sub-ledger tie-outs done and substantiation due tomorrow.
Reconciliation is where accountants say the close time goes: 55% of accounting teams name reconciliation automation as their top AI opportunity and 5% call it highly automated, and teams at the top AI maturity level close in 6.7 days against 8.7 (FloQast, 11 August 2026) 9. Cash reconciliation alone takes 20 to 50 hours a month across three to five systems (Ledge, 2025) 8.
The vendor field is the densest in the chain. Microsoft’s Account Reconciliation Agent raises sub-ledger exceptions, recommends the fix and processes within daily and monthly limits the controller sets (Microsoft, 8 September 2026) 21 22. BlackLine’s Verity Prepare adopters cut manual preparation by up to 92%; auto-certification rose from 75% to 90% and auto-match from 66% to 88% (BlackLine, 27 July and 4 August 2026) 25 27.
SAP, Oracle, Workday, Trintech, FloQast and OneStream name reconciliation or close agents of their own 14 20 33 34 36 43. The customer numbers are mostly vendor-published: Konica Minolta automated close to 99% of 45,000 monthly bank line items, and ConferenceDirect cut its close from 10 days to under 5 (HighRadius, 1 April 2025) 37. Gartner expects cloud ERP with embedded AI to deliver a 30% faster close by 2028 (Gartner, 24 February 2026) 1.
For a manufacturer the hardest accounts are the supply chain ones: the goods-receipt clearing account, stock at standard cost against the warehouse count, consigned and subcontract stock, the intercompany pair in two ERPs. Their sources sit in materials management and the warehouse system as much as in the ledger, which is why data read in place and one ontology for the purchase order line and the receipt matter most here.
From the last working day to workday 5, general ledger accountants in the shared service centre export sub-ledger balances (payables, receivables, fixed assets, inventory, the goods-receipt clearing account, payroll, tax) and ledger balances from each ERP into one spreadsheet per account. They download bank statements from each portal, match lines to cash postings and chase the rest.
For every difference someone hunts the supporting document, writes an explanation and prepares an adjusting journal for approval. A senior accountant reviews, a controller certifies, and the close manager updates a checklist by e-mail and in a tracker. In an estate with two ERPs the same account exists twice under different numbering, and the group view is built by hand.
The elapsed time is most of workdays 1 to 5 for the general ledger team, and the tail of aged, unexplained items runs for months. Auditors sample the certifications long after the period closed.
Left: the close week as the note describes it. Right: the designed path, with the same people deciding the differences. Showcase estate.
The six stages are the ones chapter 6 of the story set out; here they carry this use case’s objects, questions, terms, trigger and measure. Everything below is a design: the roles would be configured from the same parts as the PO cancellation agent, and the finance objects (ledger account, open item, reconciliation item, close task) would be added as an extension of the ontology.
| Row | Kind | What it tests or does | Data it reads | Then |
|---|---|---|---|---|
| 01 | Check | Ledger balance against its sub-ledger, bank or clearing source | ledger balance, open items, bank lines, clearing items | carry on |
| 02 | Check | Difference inside $1,000 and every item under 30 days | difference, item age | yes: row 04 · else: row 03 |
| 03 | Action | Draft the explanation and the adjusting journal | documents found, prior explanations, account owner | the accountant’s queue |
| 04 | Action | Certify with the evidence attached; update the close task | reconciliation item, close task | record kept |
| 05 | Limit | Write-off value against $25,000; daily posting budget 200 | journal value, run counter | two named people · hold |
The policy sentence a controller would state, and the planning table the Planner Agent would draft from it. Designed, yet to be built; the thresholds are illustrative.
The market runs this as draft-and-approve, with execute-within-rules for matching and low-risk certification: Microsoft’s agent processes within limits the controller sets 21 22, BlackLine keeps final authority with the accountant while auto-certification reaches 90% at adopters 25 27. The blueprint designs the lead role at A3, certification and escalation at A2, and the bank and sub-ledger tie-outs at A4: one step above where most buyers start. A first release would run at A2; accounts inside tolerance would move to A3 on the recorded cases, as the PO cancellation agent’s did.
| Element | Setting | Set by | Changes |
|---|---|---|---|
| Objective | Tie every balance-sheet account to its source each day; certify inside tolerance; draft the rest for a person | J. Harrison, finance lead | New version |
| Data scope | Ledger and sub-ledger balances, open items, bank lines, clearing items, intercompany balances and close tasks in SAP S/4HANA and the Oracle Fusion division | A. Novak, data lead | New version |
| Rules | The reconciliation policy and its planning table: inside $1,000 and under 30 days certifies; the rest is drafted | J. Harrison with the Planner Agent | New version, simulated first |
| Trigger | Daily after the overnight postings and on each bank statement · the close calendar, workdays 1 to 4 · on demand per account | J. Harrison, finance lead | At once |
| Actions permitted | Post an adjusting or reclassification journal · clear an open item · certify with evidence · update a close task · open an escalation | J. Harrison, owner | New version |
| Limits | Tolerance $1,000 · write-off under the rule $500 · daily posting budget 200 · confidence floor 0.85 · two named people on write-offs of $25,000 and above | J. Harrison with the CFO | New version |
| Escalation | Above tolerance, to the account’s accountant · open past 5 workdays, to J. Harrison · a failed posting, to the owner | J. Harrison, owner | At once |
| Level | A2 Supervised with the account’s named accountant as approver; A3 for accounts inside tolerance once the recorded cases pass | The CFO | Up on evidence, down in one step |
| Feedback | Every accept, edit and reject on a draft is recorded and used in the next cycle | Always on | Fixed |
Every row is a setting or a versioned rule; the level and the limits are the ones a manager signs.
Lima’s agent blueprint is a design of 960 agent roles across five end-to-end chains. It is the map of what Juno can be configured to do, and the first agents are built from it one at a time, the way the PO cancellation agent was.
Additional capability in Record-to-Report lists the rest of the map for this chain.
| Measure |
|---|
| days to close (baseline: 59% of companies within six business days; 57% of accounting teams at seven days or more) |
| share of reconciliations certified under the rule (adopters: 75% to 90%) |
| auto-match rate (adopters: 66% to 88%) |
| preparation time per reconciliation (adopters: 92 to 94% lower) |
| hours on cash reconciliation (baseline 20 to 50 a month) |
| unexplained items older than 60 days, count and value |
| adjusting journals per close and reversals after certification |
Baseline days to close first, because it is the number the CFO already tracks and the one the market states most often. Count the workdays from the last day of the period to the certified ledger for the last six closes, and beside it the share of accounts a person certified inside tolerance, the work the rule would take over. Those two lines make the case for a level review after the first quarter.
At A2 a named accountant approves every journal before it posts. At A3, after the recorded cases pass, the entries inside tolerance would post under the rule and everything else would wait for a person.
The mapping links the two account numbers to one ledger account in the ontology, so the tie-out reads both in place and the group view comes from one query.
The reconciliation record: the rule version, the balances as read, the documents found, the decision by name and the ERP’s document number. Internal audit reads a sample of it each quarter, the way chapter 7 reads the PO agent’s record.
The tie-out finds a difference after the journal that caused it has posted. Could that journal have been held before it reached the ledger?
Every journal scored as it posts, the unusual ones held with a reason, and the material ones waiting for a named person.
Bring two systems and one decision your planners make every week. In the first week we register the data, confirm the mapping with your team, answer your questions in ChatLima and run the first agent at the level you choose.
