- You share your Tally data
We start from the ledgers you already keep: sales, purchase, stock, and receivables and payables registers. Today that means a Tally export; a direct Tally connection is being built so updates flow without manual exports. Around 24–36 months of history works best, and seasonal demand forecasts need at least 24 months.
- We check whether the data can be trusted
Before any model runs, data-quality checks look for gaps, duplicates and entries that don't add up. You get a fix list for your accountant, and weak data is flagged rather than hidden. The full Data Trust score, with GST and bank reconciliation, is in development.
- Four Foresight™ models build your baseline
SupplyChain forecasts demand and reorder points. Finance finds trapped working capital, projects cash gaps and flags unusual ledger entries. Revenue forecasts the order book and Customer scores concentration and default risk. Each uses a method suited to its own data, documented in our AI Model Card.
- You get a readout, not a dashboard
Each readout shows the forecast, the exceptions worth your attention (named SKUs, ledger entries, accounts) and a specific recommended action for each. Nothing is executed automatically: you or your finance lead decide.
- Preshocks test what could break In development
Scenario Foresight will run named preshocks, such as collections drift, a big buyer paying late or a bank cutting your limit, against your baseline. Each run shows when a cash gap would open, your runway, DSCR and drawing-power headroom.
What we need from you
- Tally exports of your sales, purchase, stock and receivables and payables registers.
- A finance or accounts contact who can answer questions about the books.
- For Scenario Foresight, when it launches: your bank sanction letter (limits, drawing-power margins, covenants), GSTR-2B data and a list of MSME-registered suppliers.
How long it takes
The design target is about two weeks from receiving your data to a first readout, because there is no new ERP to install. The exact timeline depends on how clean the data is and how much history you have, and we agree it during the discovery call.
What sits underneath
BlackSwan Foresight is built as an agentic system using the Model Context Protocol (MCP): a set of specialised tools, each doing one job, coordinated by an agent that runs them in the right order for your request. That keeps each step explainable and auditable, instead of one opaque model.