Stop forecasting in spreadsheets. It's wrong and you know it.
A 60%-accurate forecast becomes 88-92% with Kevin. Idle balances turn into yield. Pix (Brazil's real-time payment system) 24/7 becomes intraday management. All across the 40+ banks you already operate.
A 14% Selic rate changes everything.
Text
Every R$ 1M sitting in a checking account for 30 days loses R$ 12.5k in yield. Pix 24/7/365 killed the "end of day." An Excel forecast lands at 60% accuracy, and a bad forecast pushes you into emergency credit at Selic + 10 p.p.
- Spreadsheet forecast: 60% accuracy
- Agentic forecast: 88-92% over 13 weeks
- Automated sweep: +25-50% in liquidity optimization
- Bank fee analysis: 5-15% savings (Trovata case: US$ 700k/year)
Four fronts.
Front 1 — Multi-horizon forecast
13 weeks, 12 months, scenarios (base, upside, stress). Learns your idiosyncratic patterns, explains variances, answers ad-hoc questions.
Front 2 — Automated sweep
Allocation into CDBs, repos, and Tesouro Selic (Brazilian government bonds) within your liquidity policy. Counterparty limits. Optimized maturity curve.
Front 3 — Intraday Pix 24/7
Payment prioritization under stress. Overdraft prevention. Contractual × reputational × regulatory optimization.
Front 4 — Bank fee analysis
Fee mapping across every bank. Non-standard naming normalized. Benchmarking and renegotiation.
Kevin's math at a R$ 500M company.
Projected annual gains
- Incremental yield from sweep: R$ 3.5-4.5M
- Fee savings: R$ 100-750k
- Avoided emergency credit interest: R$ 500k-2M
- Fraud prevention: R$ 100-500k
- Total projected: R$ 4-7M/year
Kobana Team Business cost: R$ 90k/year. Payback: ~8 weeks.
Services holding with R$ 1.2B in revenue, 12 entities.
Before
- 13-week forecast in Excel, ~55% accuracy
- Weekly sweep, subjective, conservative
- Reconciliation 5 days after the movement
- Bank fees never audited
After (12 months with Kevin)
- Agentic forecast, 91% accuracy
- Daily, automated sweep
- Real-time reconciliation
- R$ 680k/year in renegotiated fees
- R$ 6.2M/year in incremental yield
- 2 BEC fraud attempts blocked (R$ 340k protected)
Stack involved
List
- Kevin (lead agent)
- Klaudinho (Q&A and consolidation)
- Open Finance across 40+ banks
- CNAB 240/400 (Brazilian bank file formats)
- Pix API real-time
- B3 / BM&F
- PTAX Bacen
- ERPs (SAP, Oracle, TOTVS)
FAQ
Does it work for a seasonal business?+
Yes. The model learns seasonality (Black Friday, harvest, back-to-school, tourism).
Can I configure a custom liquidity policy?+
Yes, all through Kia Brain. Counterparty limits, allowed product types, redemption windows.
Does it replace a TMS (Kyriba, SAP Treasury)?+
It complements. Runs on top as an agentic layer.