KINTO replaced a two-decade-old system and a macro-driven spreadsheet with an agentic ERP that matched the old system to the cent on 100% of shared waybills over six months and surfaced customer-reactivation opportunities.
Problem
The operation ran on a web application more than twenty years old, with obsolete dependencies and hundreds of thousands of waybills in one database. The twice-monthly financial close lived in a macro-driven spreadsheet, and every day someone retyped data that already existed on a third party’s platform.
Signals
- Market: The company works on the platform of a third party that settles its earnings, so keeping each waybill’s accounts right per customer is the business.
- Business: Each customer’s activity was not visible in time: customers who stopped shipping went unnoticed.
- User experience: Dozens of screens per task and real operational errors (files downloading with broken names).
- Support: Receipts and waybill statements are sent over WhatsApp and email.
How we understood it
- Legacy system audit: inventory of tables and volumes.
- Recorded, transcribed meetings (one of nearly four hours) turned into a prioritized list, each item with its timestamp and a verification in code.
- Authorized analysis of the third party’s platform: its real limits were documented.
- Waybill-by-waybill cross-validation, by waybill number rather than totals, across six months of billing periods, plus an economic valuation of the ERP with every figure labeled measured, market or estimated.
What we built and why
- Rescue without a rewrite: the old system was stabilized and redeployed, bulk loading was rewritten (from timeouts to a fixed number of queries per batch), and automated tests were added.
- Agentic ERP replacing the application and the spreadsheet: automatic ingestion from that platform several times a day, full history loaded behind verification gates, a single customer master, a financial engine (daily consolidation, settlement with configurable taxes, billing statements, receivables, treasury) and a compatible facade so a large customer’s integration kept working.
- Labeling tool for sticker printers: a waybill of N boxes produces N labels with a valid barcode in one click, licensable per customer, with no content leaving the browser.
- Public site with an SEO/GEO layer, structured data and conversion tracking.
Why this way: The guiding principle was “autonomous by default, human by exception”: the new system handles the routine on its own and calls a person only when something falls outside what is expected.
Transformation
-
Six months of billing periods matched the old system to the cent on 100% of shared waybills; the remaining differences were explained one by one. Receivables reconciled customer by customer.
-
New visibility: customer-reactivation opportunities among high-value “dormant” customers, detected by the system.
-
Active production with the old system frozen.
-
historical waybills loaded: hundreds of thousands (more than a decade) (measured)
-
reconciliation against the old system: 100% of shared waybills, 6 months (measured)
-
freight savings from the semi-bulk load rule: ≈29–50% depending on tonnage (measured)
What we don’t know yet
- Data-entry hours saved: the valuation itself calls for timing them.
- A pay-for-results pricing model was analyzed, with no agreement closed to date.
