The platform
A bag of feed leaving the store is a stock movement and a journal entry, posted once. That single decision is why a cost per kilo of gain, a cost per egg and a set of statements a bank will accept all come out of the same system — instead of out of a spreadsheet somebody rebuilds every quarter.
Component
Today
The data path
Farm data does not arrive clean. It arrives as a photographed notebook page at six in the evening, from someone with no laptop and one bar of signal. These steps are numbered because the sequence matters — post before you read and the ledger is wrong, report before you post and the report is fiction.
A tally sheet at the shed door. This is where almost every farm record in the region actually begins, and a system that pretends otherwise is a system nobody fills in.
Excel, PDF, or a photograph of handwriting. Hedge reads the document, maps the columns and proposes the records — it does not ask the shed to retype anything.
Guards live in the database, not just the form: head counts that cannot go negative, a cumulative total that cannot be logged as a day's figure, an edit that has to leave a reason.
The operational record and the journal entry are the same event. There is no nightly sync to go wrong, because there is nothing to sync.
A statement pack, a management report, an export under your own letterhead — and every figure on it drills back to the rows that made it.
Modules
Each one is useful alone. The reason they are worth having together is that a figure from any of them can be traced to a figure in any other without a person in the middle re-keying it.
Livestock
Parentage, weight log, average daily gain and feed conversion for every animal. Breeding calendar and farrowings, automatic stage transitions as animals age, and mortality that reconciles against head count instead of drifting from it.
Land
Blocks and plots, planting through harvest, input applications that draw stock down as they are recorded, and yield booked against the land that produced it.
Inputs
Ration formulas over an ingredient library, procurement and deliveries, and issues to a named house or pen — so feed consumed can be set against the weight it bought.
Finance
A proper chart of accounts and double-entry journal behind every record, with an IFRS-aligned statement pack, budget versus actual, AR and AP aging, bank and capital reconciliation, and an inter-farm account between sites.
People
Payroll runs that write to the ledger on approval, so labour lands in the cost of the batch that used it rather than in a lump at month end.
Reporting
Management reports and a full statement pack, exported to PDF, Excel or CSV under your own letterhead — with every figure drilling through to the records behind it.
Automation
Where the machine helps is at the edges of the record — getting it in, getting it out, and refusing an entry that cannot be true. Not in deciding what the farm should do.
The assistant finds the figure, opens the screen it came from and shows the rows behind it — so an answer can always be checked, never just trusted.
Daily sheets arrive as Excel, as PDF, as a photograph of handwritten paper. Hedge reads them, maps the columns and files the records.
Task digests by email, and alerts by SMS and WhatsApp, so a vaccination due or a mortality spike reaches the person who can act on it.
Guards in the database, not just the form: head counts that cannot go negative, cumulative totals that cannot be logged as a day's figure, edits that leave a reason and an audit trail.
Reading a photographed tally sheet, mapping an unfamiliar spreadsheet column, and answering a question in plain language — that is the whole remit. Every answer opens the screen and the rows it came from. We publish no claim that AI raises output, because we went looking for the evidence and did not find it; what we found instead is on the evidence page.
Control and audit
A record only has value to a bank, an auditor or a buyer if it cannot be quietly rewritten. These are the controls that make the difference between a database and a book of account.
Permissions are enforced row by row in the database, so a role that should not see payroll cannot reach it by any route — not through a report, not through an export, not through a URL somebody was sent.
Payroll runs, journals and adjustments wait for approval before they post. The person who raises an entry and the person who approves it are recorded separately.
Changes to a posted record are not silent. The system asks why, keeps the before and the after, and stamps who and when — which is what makes the trail worth reading a year later.
Constraints live in the database. A head count cannot go negative, a cumulative total cannot be filed as a daily figure, and a duplicate day cannot be inserted — regardless of which screen, import or script is trying.
Tenant isolation is enforced at the row level in the database rather than by a filter in application code, because a filter somebody forgets to write is a data leak, not a bug.
Repeated failed sign-ins lock the account, and an idle session ends itself — because a shed office computer is not a private desk.
Built for the field
Field conditions
Design constraint Software written for a farm in Iowa assumes the connectivity, the currency and the paperwork of a farm in Iowa. Hedge was written on a farm in Rwanda, against the problems that farm actually has. Hedge · Kigali, Rwanda