How to Automate Jirav

Every department is spending against a number set in Jirav

Jirav is the forecasting, budgeting, reporting and dashboarding tool that accounting and CFO advisory firms use on behalf of their clients, and that finance teams inside small and mid-sized companies use for themselves.

What it holds is not the past. The ledger already has the past. Jirav holds the plan: what revenue is supposed to be in November, how many people should have been hired by then, what gross margin the board was told to expect, and how much cash is left at the end of it.

That makes it a different sort of document from a set of accounts. A marketing director spends against a number in it. A hiring manager opens a role because a line in it says the role exists. A lender reads a covenant forecast that assumes it. When the plan is wrong, everybody downstream is wrong in the same direction.

Half the model is numbers nobody in finance owns

The accounting figures look after themselves. The assumptions do not.

A driver-based model runs on customers, units, headcount, utilisation and price, and almost none of those live in the general ledger. The pipeline sits in the CRM and the sales director updates it on Thursdays. Open roles and start dates sit in the recruiting system. Churn is a report somebody runs. The cost per acquisition the growth driver assumes comes off an ad platform's dashboard. The input price the gross margin depends on is on a supplier's portal and moved in April. The rent escalation is in a lease PDF, and the rate on the credit facility is on the lender's own site.

So the model gets built carefully in January and then patched from memory, one figure at a time, in the two days before a board meeting. And it usually belongs to one person. In the week they are away, the plan stops moving and everybody keeps spending against it.

Jirav

Drivers refreshed on the model's schedule Pipeline, headcount, churn and input prices read from the systems that hold them and put beside the forecast lines that assume them.

Budget drift caught mid-month Departments tracking ahead of plan flagged with two weeks left to act, instead of appearing as a variance once the period has closed.

Board prep that starts assembled Actuals, variances and the metrics a board always asks about gathered into a working file before anyone opens the deck template.

Jirav rebuilds the three statements from last night's actuals

Jirav handles the modelling side well, and a team that has only ever built a budget in it is using a fraction of it.

Driver-Based Financial Modeling is the core, with forecasts built from operational drivers rather than a percentage added to last year. 3 Statement Modeling keeps the profit and loss, the balance sheet and the cash flow moving together, so a change to a hiring plan shows up in the runway and not only in payroll. Workforce Planning forecasts new hires, raises, bonuses and accrued expenses, and driver-based hires let you say one customer success manager for every fifty customers rather than naming each one. Scenarios run what-if versions across all three statements, and rolling forecasts adjust as actuals land. Underneath it, native connections to QuickBooks Online, Xero, NetSuite and Sage Intacct import the trial balance each night, with payroll systems such as Gusto and ADP alongside.

The ceiling is exactly where that list stops. Everything Jirav syncs on its own is an accounting or payroll number. Every driver that is not one gets typed in by a person, and the model cannot know when that person last checked. A forecast built on a nine-week-old assumption recalculates to four decimal places and is still wrong.

A model whose assumptions are checked as often as the actuals

The actuals refresh overnight. There is no reason the drivers cannot.

Weighted pipeline read out of the CRM every Monday and put beside the revenue driver, so the forecast reflects what sales believes this week. Open roles and accepted offers taken from the recruiting system and compared against the hiring plan, so a plan that assumed six engineers by June is corrected in February. Spend and cost per acquisition pulled from each ad platform, so the growth assumption is a measured number.

Then the prices. The supplier's portal checked for the input cost the margin depends on. The published index a contract is tied to. The rate on the credit facility read on the lender's site rather than remembered. Utilisation and billable hours taken from the practice system, so a services forecast stops waiting on a monthly export.

And the model's own hygiene. Which models have not been refreshed since the last close. Departments tracking ahead of budget with two weeks of the month still to run. Drivers outside their normal range, flagged the morning they move rather than at the next quarterly review.

The forecast stays the finance team's opinion

None of that decides what the plan should be. It changes how current the inputs are when a person decides.

WebRun is an AI agent that works a real Chrome browser, signed in as you. It opens the CRM, the recruiting system, the ad platform, the supplier portal or the bank, reads the figure, and puts it where your model expects it, on the morning you need it rather than the week after.

It runs on your schedule in your own private environment, and each workflow is limited to the sites you name. Anything that leaves the building is drafted and waits. A board pack or a lender package is a number somebody signs their name to, and the signing stays with them.

The workflows below are already built, and each one names what it opens.