Build your model once. Run it for every client.
Your templates, your rules, each project's context. Your own Claude does the work in a fully auditable spreadsheet, and every change stays yours to revert.
Free to start, free to keep · No credit card · No AI training on your data
Northwind's pack, in their order: ARR bridge, P&L vs budget, cash. Actuals went to the Actuals branch, the voted budget is untouched.
Layerz get_conventionsLayerz patchRunway is under Northwind’s 12-month floor, so it opens the commentary, as the project’s rules ask.
Two lines explain all of it, traced to their entries:
Layerz readLayerz list_transactionsHosting follows the usage spike since 4 March; the recruiting fee is the senior AE hire, one-off. Each gets its one-line comment on the row, as Northwind’s rules ask for any variance above 10%.
Done.
Layerz exportStandard Excel with live formulas. Nobody on the other end needs a Layerz account.
Layerz · NorthwindOpus 5.5Ten clients. Ten versions of the same model.
Every new client starts from an old file that is "close enough". Every month, actuals get pasted, a value hardens, a formula stops at column N. And every Claude session starts from zero: their chart of accounts, their fiscal year, the way their board likes a variance explained, all typed again.
Your method lives in your head and in a folder of copies. It should live in one place, and work the same for every client.
A client's year, four moments.
Start a new client in an afternoon
Open a project for the new client and fork your template. Import their last Excel, plug their books (Pennylane, Qonto, Stripe, DATEV, FEC or any export) and set the mapping once. Write the project's context: what they do, fiscal year, what the board reads first.
The monthly pack, the same way every month
Actuals land on their own branch, the voted budget never moves. Variance and reforecast compute. Claude drafts the commentary in the order and with the thresholds written in the project's context.
The board, the bank, the raise
Every number traces back to its assumption and its entry. Scenarios side by side. A clean .xlsx with live formulas, or a share link nobody needs an account for.
The handover
When the mandate ends, the client keeps a live, documented model, or you hand them the whole project. No file nobody understands.
Your method, written once. Each client its own project.
Your rules
Signs, controls, how you explain a variance: written once, applied in every project.
Each project's context
On the first screen of the project, read by Claude before it touches a number.
Your templates
Your monthly pack, your board pack, your fundraising plan. Publish them to your library, fork them for each new project.
Never mixed
Each project lives in its own workspace. Someone you invite on a model sees that model and you, nobody else.
A grid you already know. Fully auditable.
Named lines instead of cell references, one formula per row, a balance check that stays at zero. Type in a cell or ask Claude: same live model. Export to Excel whenever a client wants the file.
Three models fractional CFOs run every month.
Monthly close & client reporting
Budget vs actuals, reconciliation of ledger, sales journal and payroll, close points, next year's plan.
SaaS board reporting & forecast
ARR, unit economics, P&L, cash and balance sheet, actuals and forecast.
Fundraising business plan
Subscriptions and services, team, funding and cash runway, monthly over three years.
Free to start. Free to keep.
Create it and work with it for real: unlimited models, client work included, your own AI, clean .xlsx export. No time limit.
TeamWhen each project gets its own workspace with its own context and templates, when a colleague or a client edits with you, and for a longer version history.
Several CFOs, one method.
When the method lives in people, every CFO rebuilds the same model their own way. Put the firm's templates and rules in one library, give every mandate its own project, and every client gets the same quality, whoever runs the mandate.
“Don't make the mistake of asking an LLM to calculate. You have the right approach: bring AI into a framed infrastructure.”
Can a client see my other clients?
No. Each client is a project in its own workspace. Someone you invite on a model sees that model and you, nobody else.
My clients want Excel.
They get it. Every model exports to a standard .xlsx with live formulas, on every plan. Nobody on their side needs Layerz.
Does Claude do the math?
No. Claude writes the structure, the Layerz engine does every calculation, the same way every time. Every number traces back to a formula you can open and question.
What happens to the model when my mandate ends?
The client keeps it: a live model they can open, audit and export, or the whole project handed over to them.
What do I need to start?
The AI you already use (Claude, ChatGPT, Codex or any MCP client) and a couple of minutes to add Layerz, free plans included. No credit card for Layerz.
Every client. Every close.
Everything reconciles.
Free to start, free to keep · .xlsx export always free · Works with any AI