For the office of the CFO · 40.71°N 74.00°W

Why is the number different this time? Gartner asked 497 accountants how often they get something wrong. 59% said several times a month — from manual work, from reopening the books, from a review too rushed to catch it. Every one of those is a property of rebuilding the same schedule by hand, every quarter, against the clock. Gartner, February 2024 — survey of 497 accountants

The exposure review that took a week —
now runs before the close.

Locra reads the exports your ERP already produces and builds the reviews finance repeats every quarter: where you have crossed a tax threshold, where revenue is concentrated, where the cash is stuck. Same inputs, same numbers, every run — as a PDF or Excel pack that lands on schedule, built on your machines, with every step open to read.

Launch news — nothing else.

  • Reads your ERP exports in place
  • Deterministic — the same run gives the same number
  • Your ledger never leaves the building
  • Scheduled — the pack arrives before the close

Your quarter today

  • A threshold table per state, and a spreadsheet marathon to compare against it
  • Exposure that triggers on order count, invisible to anything sorted by revenue
  • “Who are our biggest customers?” answered; “where are they?” never asked
  • The AR aging export filtered six ways and pasted into six emails
  • Every one of these rebuilt by hand, every quarter, by the person you can least spare

Every one of these is a demo below — built once, scheduled, delivered.

Built from real finance workflows · 40.71°N 74.00°W

Three jobs, done end to end

Each demo is a complete job — from the ERP export to the delivered pack — built on a realistic mid-market ledger: one fictional manufacturer, ~$4.7M of trailing-12-month revenue, 638 customers, selling into 43 US states and 8 countries. Same two files, three different questions.

revenue by state vs threshold $100k CA NY IL NC OH 232 txns triggered on count, not revenue

Nexus & tax exposure by jurisdiction

Flagship demo

Where are we paying tax because of where we do business? Trailing-12-month revenue and order count per state, compared against a per-state threshold table. Eight states triggered — including one at $67k of revenue and 232 orders, which no spreadsheet sorted by revenue would ever surface.

An exposure map, triggered and approaching states, and the customers behind each — as a scheduled pack.

top 10 32% of revenue top 2 16% of revenue

Revenue concentration, on a map

Two-minute demo

Every CFO can name the top customer. Almost none can show where revenue is concentrated — by metro, by region, by dependency. Top ten customers: 32% of revenue. Two of them: 16%. A board question and a sales question, from two files, in under three minutes.

A revenue map, revenue by state and country, and a cumulative concentration table.

aging cur 90+ $205k of $694k overdue US-SOUTH 1 / 6 one per region

Collections pack by region

Automation demo

$205k of $694k overdue — and the aging export won't tell you that one region has a payment-culture problem spread across many accounts while another is a single German customer holding 91% of its overdue. Same number, two completely different actions.

One pack per regional manager, emailed on the schedule; the consolidated view and overdue map to the CFO.

Recorded runs are on their way — leave your email and we'll send each one as it lands.

The half of finance nobody maps · 40.71°N 74.00°W

Everything that happens in business happens somewhere

Almost every number the office of the CFO cares about has a location attached to it, and almost no finance tool can use it. Locra treats where as a first-class column — the maps, the thresholds-by-jurisdiction and the spatial joins are in the box, not in a separate department.

Where are we creating tax liability?

Revenue and transaction counts rolled up by jurisdiction, measured against the rule that applies there — with the states you are quietly approaching flagged before they trigger.

Where does our revenue actually come from?

Customer revenue on a map, by metro, region and country — so concentration risk is something the board can see rather than something the CFO asserts.

Where is our cash stuck?

Overdue balance mapped by customer location, aggregated by region — separating a regional payment culture from one large account that needs a phone call.

Where does the estate cost us most?

Cost per seat, tax value and operating cost across a property portfolio, compared building by building — the consolidation question, computed instead of argued.

Built for a controls conversation

Finance cannot sign off on an answer it cannot reproduce. Ask a chatbot the same question three times and you can get three numbers; that is fine for a first draft and unacceptable for anything anyone signs. Locra uses AI to build the workflow — and then the workflow, not the model, produces the number.

Same inputs, same number

Once built, the workflow is ordinary code running on your data. Re-run it a hundred times and it returns the same answer a hundred times — the model is not in the loop any more.

Every step is readable

The AI writes a plan you approve, then real Python you can open and check — step by step, node by node. Nothing about the calculation is hidden inside a prompt.

Your ledger never leaves the building

Invoices, customers and balances are read where they live and never uploaded. No rows of your financial data are sent to any model, at build time or after.

Flags for review, not advice

The exposure pack tells you which jurisdictions to review with your tax advisor and shows the evidence behind each flag. It is built to support the judgement, not to replace it.

1× you pay the AI once — to build the workflow
$0 every quarter it runs after that, forever
0 rows of your ledger ever sent to the model

Fits the stack you already run

Reads your exports where they are

The invoice export, the customer master, the AR aging, the threshold table you maintain by hand — spreadsheets, CSVs and databases, read in place. No migration, no new system of record.

Excel out, because that's the ask

PDF for the board pack, Excel for the person who is going to work in it anyway. “Can I have this in Excel?” is a real sentence, and it gets a real answer.

Scheduled — before the close, not after

Put the review on a schedule and the pack lands where you expect it, on the day you need it. Scheduling is included, not a server product sold separately.

Yours to re-run, free

AI builds the workflow once; every re-run after that is yours, forever, at no per-run cost — and every step stays open to read and change.

See it on your own ledger

Leave your email and we'll send the recorded demo runs at launch — and when you're ready, bring one invoice export and watch your own exposure review build itself.