Openi Artificial Intelligence
Expand what your teamcan get done.
AI agents that run finance and accounting routines, investigate open items and prepare deliverables with the context of your operation.
Openi agents
3 routines running
Finance
Reconcile October statements
1,271 reconciled2 to review
Accounting
Categorize entries with no rule
38 suggestedfrom history
Tax
Read and check incoming invoices
112 invoices linked1 with no payment
Sample data.
01 / 07 · Request
Illustrative sequenceNew task
Company
Grupo Vértice
Branch
Branch 03
Period
October 2026
Routine
Monthly review
Request
Company and period selected.
The review starts with the company, the branch and the month. The rest of the context comes from how your operation is set up.
Sample data.
- Request: Company and period selected. The review starts with the company, the branch and the month. The rest of the context comes from how your operation is set up.
- Start: The review routine starts. The steps are your routine's: read statements, match against the ERP, check invoices, set aside what doesn't match.
- Sources: Records and documents checked. Bank statements, the ERP ledger and captured invoices, each source with what it brought in.
- Mismatch: A debit with no entry. An Itaú debit has no match in the ledger. It leaves the flow and becomes an open item.
- Recommendation: A suggestion with evidence. The open bill for the same amount and the supplier's invoice appear next to the suggestion.
- Review: Sent to your team. The suggestion waits for a decision from someone who knows the operation. Nothing goes to the ERP before that.
- Summary: The work, summarized. What was read, reconciled, linked and set aside, and what is still waiting on someone.
In action
Meet your nextway of working.
One task moving through the platform: from the request to the summary of what was done.
Agents
AI connected to the workyour team does.
01
Works with context.
Data, documents, rules and history on hand to guide each task.
02
Runs the routine's steps.
From checking the sources to preparing the result, with a record of what was done.
03
Brings your team into the decision.
Open items and recommendations arrive with what's needed to review them.
Execution
Put the routinein motion.
The AI works alongside Openi's engines and rules: what has a rule follows the rule, what has an exact match the engine reconciles, and the AI steps in where context has to be read.
Input
Sources compared
- Itaú statement · Oct642 transactions
- Bradesco statement · Oct642 transactions
- ERP ledger · Oct1,279 entries
Actions
- Reconciliation engineMatches amount, date and reference
- AISuggests a link with no exact match, from history and the counterparty
- Platform workflowSets aside what doesn't match
Output
Matches and open items
- Reconciled1,271
- Suggested links11
- Open items to review2
Sample data.
ReconciliationInvestigation
The difference shows up.So does the investigation.
Pull together the related records, check the context available and prepare the next step to resolve the open item.
Selected record
Debit with no matching entry
-R$3,912.40
- Account
- Itaú · account 4471-2
- Date
- 10/14/2026
- Memo
- SISPAG FORNEC 0981
Context checked
SearchingERP ledger
No entry for this amount between Oct 10 and 18
Accounts payable
Bill 0981 open · Lince Transportes · R$3,912.40 · due Oct 15
Captured invoices
NF-e 22814 from Lince Transportes, issued Oct 2, same amount
Proof of payment
Not found in the documents received
Next step
Possible payment of bill 0981.
Amount, supplier and reference match. The payment date and the proof of payment were not confirmed.
Suggest settling bill 0981 after review
Sample data.
Amount mismatches
The same payment with different amounts on each side.
No match
A record on one side with no counterpart on the other.
Possible duplicates
Two records that look like the same event.
Missing information
An invoice, receipt or detail still needed to close it out.
Categorization or adjustment
The suggested account or correction to review.
The investigation gathers what exists. When the evidence isn't enough, the item ends in a request, not a conclusion.
Analysis
Take the analysis down to the datathat explains the result.
Compare periods, find where spending concentrates and drill into each number down to the entries behind it, on data that has already been reconciled and categorized.
Use cases
- Compare periods
- Find where spending concentrates
- Investigate changes
- Sort open items by company
- Prepare summaries and reports
Analysis taskWhich expenses contributed most to this month's change?
- Period
- Oct vs. Sep 2026
- Company
- Grupo Vértice · Branch 03
- Sources
- Itaú, Bradesco and ERP
- Updated
- Nov 6 · 8:40 AM
Operating expenses rose R$48.2K over September. Freight and IT services account for 62% of that change.
Contribution to the change
Total change+R$48.2K
Entries used · Freight
- Oct 3Lince Transportes · Invoice 22701-R$9,840.00
- Oct 14Lince Transportes · Invoice 22814-R$3,912.40
- Oct 22Rota Sul Cargas · Invoice 5120-R$4,647.60
The analysis shows where the change is. The reason is confirmed by someone who knows the operation.
Sample data.
Comparisons and drill-down to the entry use Openi's reports. The sample data is for this example page.
Reporting and analyticsInstructions and context
Your criteriaguide the work.
Execution follows the structure and rules of your operation. Each criterion changes what happens to a record.
Operating criteria
- Company and period
- Grupo Vértice · Branch 03 · October
- Chart of accounts and cost centers
- The company's chart · 14 cost centers
- Rules and procedures
- Fees to Bank charges; freight by invoice
- History available
- Entries approved by the team
- Authorized systems
- Itaú, Bradesco, ERP and captured invoices
- Approval conditions
- Adjustments are reviewed before the ERP
Instruction
“Set aside items without documentation and send adjustments for review before they go to the ERP.”
Effect on execution
- Freight · Lince Transportes-R$9,840.00Under review
- SISPAG debit 0981-R$3,912.40Under review
- Itaú fee · recategorization-R$48.90Under review
- Support · Nuvem Clara-R$3,000.00Under review
The adjustment waits for approval before it's sent. The item with no document stays out of the batch until someone completes it.
Sample data.
Criteria, rules and approval conditions are set during onboarding, with the Openi team.
Tracking and control
See what was done.Decide what happens next.
Tasks, sources checked, actions taken and suggestions prepared all sit in one place. What goes to the destination system passes through your team's decision.
Tasks
- October review · Branch 03Awaiting decision
- Categorization · Branch 01In progress
- Incoming invoices · week 44Waiting for data
- Reconciliation · Branch 02Done
Output
Recategorize Itaú fee from Oct 14
FromOther expenses
ToBank charges
Amount-R$48.90
- Prepared
- Approved
- Sent
- Confirmed
The suggestion is ready, with its evidence.
Sample data.
Why this suggestion
- Same fee posted to Bank charges in September and AugustSource: Entries 8812 and 7930
- Operating rule: bank fees go to Bank chargesSource: Rule 12
- Debit on the Oct 14 Itaú statement, no invoiceSource: Statement · line 311
Sources checked
- Itaú statement · Oct
- ERP ledger · Oct
- Operating rules
Actions taken
- Read 642 Itaú transactions
- Matched against the ERP ledger
- Prepared 1 adjustment
- 01Prepared
- The suggestion is ready, with its evidence.
- 02Approved
- A person on your team decided.
- 03Sent
- It went to the destination system.
- 04Confirmed
- The destination recorded it, as the integration allows.
Permissions by company and by action.
You decide which companies the AI can see and what it can do in each one. Viewing isn't editing, and editing isn't deleting.
Security and data handling (Portuguese)| Company | View | Edit | Delete |
|---|---|---|---|
| Branch 01 | Allowed | Allowed | Not allowed |
| Branch 02 | Allowed | Not allowed | Not allowed |
| Branch 03 | Allowed | Allowed | Not allowed |
Sample data.
Integrations
Connected to the systemswhere the work happens.
The task uses what the connections bring in and delivers wherever the integration allows. What each connection can read or do depends on the system.
Open BankingReads
European bank accounts through PSD2, and Brazilian accounts, cards and investments through Brazil's Open Finance, with the account holder's consent.
Direct integrationsReads
Statements and payment returns from banks and card acquirers.
ERPs and systemsReads
Entries, bills, chart of accounts and cost centers.
Tax documentsReads
Brazilian invoices issued and received (NF-e, CT-e) and documents read with OCR.
Files, databases and webhooksReads
Spreadsheets, databases and events from other systems.
Task
October review
Reads · matches · prepares
ReconciliationDelivers
Records matched and bills settled, according to the ERP.
EntriesDelivers
Categorized and sent to the ERP after approval.
Open itemsDelivers
Set aside for your team, with the evidence.
ReportsDelivers
Cash and results on reconciled data.
Reading and writing vary by system: not every integration writes the same information it reads.
Capacity
More capacityfor what comes next.
More room in the routine to analyze, serve clients, keep up with the business and make decisions.
Results from customer stories, with AI and rules working together. These are not figures for the AI on its own.
The platform runs
- Reading statements and invoices
- Matches by rule and by amount
- Categorization and link suggestions
- Setting aside what doesn't match
- Preparing entries for the ERP
Your team decides
- Adjustments and recategorizations
- Open items without enough evidence
- The operation's criteria and rules
- Analysis of the results
- Serving clients and managers
Customer stories
See what changes when AIjoins the operation.
Operations where the AI works with the company's rules, and the team reviews what needs a decision.
Fashion retail
Rent invoices read and categorized with no typing
- PDF invoice
- OCR
- Accounting rules
- CLAE and SAP
- Task
- Read each store's rent invoices, every landlord with its own PDF layout, and post them to the books.
- How it works
- AI-powered OCR extracts landlord, amounts, due dates and period; chart-of-accounts rules categorize; entries go to CLAE and SAP.
- Team's role
- Reviews and edits what's needed; every edit is logged, ready for audit.
- Result
- Reading and categorization 100% digital, no retyping.
Proven results
- saved every month
- 100 h
- digital, from reading to categorization
- 100%
- of development in rules, tables and OCR
- 320+ h
From the Openi team
The AI reads the context before it categorizes.
“Our categorization agent also studies the context: if you made a purchase, it finds the invoice, whether by the supplier's CNPJ (Brazilian tax ID) or by what was bought.”
Openi team
FAQ
What people ask about the AI.Tasks, review and data.
Didn't find what you were looking for?
Talk to a specialist and get your answer straight from our team.
Talk to a specialistIt takes part in reconciliation (suggesting links when there's no exact match), categorization (suggesting category, account and cost center from history), reading documents with OCR and handling exceptions: it sets aside what doesn't match, gathers the evidence and recommends the adjustment.
Demo
What would your team dowith more capacity?
See Openi's AI applied to a routine from your operation.
You leave your details
It takes less than a minute.A specialist gets in touch
To understand your banks, your ERP and what takes up the most time today.You see Openi at work
The AI applied to one of your routines, with your sources and criteria.
