Skip to main content
Kobanateam
Kobana Team

Reconciliation that reads the narrative, not just the amount.

From 70% auto-match (traditional RPA) to 95-99% with our agents. Monthly close from 12 days to 3. Month-end goes back to being just another day.

Why manual reconciliation takes so long.

Text

Pix (Brazil's real-time payment system) shows up with a chaotic free-text field. A boleto (Brazilian bank-slip payment) lands as a partial payment. TED transfers come with a payer name that doesn't match. A supplier deducted a 0.15% fee and no one flagged it. A three-day-old reversal. A scheduled payment that bounced on insufficient funds.

RPA breaks on any variation in description. Analysts burn 4-6 hours a day on this alone.


Kevin reconciles inflows, Kelly reconciles outflows, Karin runs cash application.

Kevin — Bank reconciliation

  • Cognitive matching across 40+ banks
  • Interpretation of chaotic descriptions ("TED JOAO SILVA REF NF 1234")
  • Handling of reversals, partials, and unexpected fees
  • Drafts notes to the controller on exceptions

Kelly — AP reconciliation

  • Matches CNAB return files with scheduled payments
  • Handles rejections (closed accounts, insufficient funds)
  • Pix × ERP reconciliation
  • Identifies unexpected charges

Karin — Cash application

  • 90-96% touchless on received Pix and boleto
  • Matching by txid when available
  • Interprets free-text fields
  • Parses remittance advice in PDF and email

What happens next.

  • 95-99% auto-match (vs 70-80% for RPA)
  • -50% to -70% reconciliation time
  • -80% to -90% human effort
  • Monthly close 12 → 3 days
  • 20 min/account (Microsoft Treasury case: -22% AHT)

Omnichannel retailer with 40 stores + e-commerce.

Before

  • 3 full-time people on reconciliation
  • 11-day monthly close
  • Pix with free-text fields matched by hand
  • 12% of operations manual each month

After (4 months)

  • 1 part-time person supervising
  • 3-day monthly close
  • 97% auto-match
  • 2.8% of operations manual (only real exceptions)
  • Team reassigned to per-store margin analysis

Stack involved

List

  • Kevin, Kelly, Karin
  • Klaudinho (Q&A and reporting)
  • Open Finance (Pluggy, Belvo)
  • CNAB 240/400
  • Pix API
  • OFX
  • ERPs

FAQ

Does it work with card acquirers (Cielo, Stone)?+

Yes. D+1, D+30 settlement, MDR, splits — all integrated.

What about D+X card payments?+

Yes. Kevin reconciles projected × actual settlement.

Multi-currency?+

Yes, with PTAX conversion and Kia Brain handling FX policy.

Ready to get started?