Frequently asked questions

Yes, in two ways, and that’s the whole list. It remembers the mappings your team confirms, so the same line comes back already resolved next month. And it indexes each past session by meaning, so you can find an old conversation by describing it. There’s no general model quietly getting better at your business.

No. A mapping carried over from last month arrives labelled with where it came from, and someone still has to confirm it. The work gets shorter; the sign-off doesn’t get weaker. That’s what makes it safe to rely on.

No. Your corrections, articles and figures stay in your own project, and no model is trained on them. We do not use your data to improve the platform for other customers. If we ever offer that, for example for benchmarks, it will be off unless you choose to turn it on. Read Privacy Policy §5.

Keep the decisions as well as the numbers. Struktürs files each mapping your team confirms against the period it belongs to, and keeps each period’s working papers, with the commentary on why figures moved. Next month the same lines come back already resolved and labelled, and last month’s papers are there to build on. Your team spends its time on what’s actually new.

Because the decisions your team made in month one come back. The lines your team mapped last month return already resolved, labelled with where they came from. What’s left is what’s actually new: a new account, a changed policy, a movement worth a second look.

Anthropic’s Claude and Google’s Gemini, on Google Cloud. Google, which runs the AI models, does not use your requests to train them. If its automated checks flag a request as possible abuse, Google may keep that request for up to 90 days to review it. Which model runs where for your country is on the Security page, leg by leg. No model writes to your Vault without a person on your team approving it. Read the Security page.

Yes. We monitor errors, speed, cost and whether answers found their sources, for each company, so we often spot a problem before you report it. Those records hold no figures, documents or names, and only named Struktürs staff can see them. You can’t run AI responsibly without watching it.

A general chatbot starts every conversation from zero and knows nothing about your business. Struktürs keeps a signed-off record of your numbers and the decisions around them, so it can answer ‘what changed between Q2 and Q3 last year?’ because it has both quarters. And when the data isn’t good enough, it tells you, instead of writing something plausible.

Paste a trial balance into a chatbot and ask why margin fell, and you’ll get a confident answer. Next month you paste it in again, and it has forgotten last month, your chart of accounts and what the team concluded. Struktürs keeps a signed-off record of every period and the reasoning behind it, in your own Google Cloud Project, and tells you when it can’t find a source. Nothing you put in trains a model.

BI tools chart data you’ve already cleaned. Struktürs does the step before: it takes your trial balance, checks it, maps it and keeps it with the reasoning behind it. You can still chart the result anywhere you like.

The AI proposes; a person on your team decides. If the data isn’t good enough for what you asked, it stops and tells you why instead of guessing. You see the verification score, the gaps, and the sources the answer drew on.

No. Nothing your company puts into Struktürs is used to train any model: not your figures, your documents or anything your team writes. The AI features receive only a minimised prompt, or a document’s text when you ask for an article to be drafted from it. We do not use your data to improve the platform for other customers. If we ever offer that, for example for benchmarks, it will be off unless you choose to turn it on. Read Privacy Policy §5.

Export a trial balance from Xero as CSV or Excel and upload it. Struktürs reads it, maps each line to a standard chart of accounts and checks the result seven ways. A direct Xero connection is on the roadmap. Until then, an export means there’s no ledger access to grant.

For the parts that repeat, yes: the trial balance checks, the mappings, the movement review and the commentary on why things moved. Management accounts and board reporting are drafted from the numbers you signed off. Spreadsheets you use for forecasting or one-off modelling stay where they are; Struktürs works from what has happened, not what might.

No. Your chart stays as it is, and Struktürs maps it to a standard one. You see each mapping, confirm it, and it’s reused next month. Over time it shows you where your chart is causing noise.

Knowledge Scaler captures what your team learns: retrospectives, risk registers, plans, ideas and signal post-mortems, each filed by period. You start a capture from Ask Vault, or from any figure in the app. The AI asks the follow-up questions and drafts the article, and your team reviews each one before it’s filed.

Once your trial balance is signed off, your team can add the context around it: sales volumes, pricing, headcount, goals, retrospectives and risks. Uploading non-financial data directly is still being built with our first customers.

Every trial balance, statement, signal post-mortem, Framework session, article and note your team has reviewed and signed off. All of it is filed by period, versioned and searchable, and exported on request.

Yes. The next FD inherits the files and the reasoning behind each period and decision, and can ask about any of it.

Ask about your business in plain English, such as ‘show me Q3’s income statement against Q2’, and get an answer with its sources and how confident it is. It only draws on your signed-off numbers and your team’s articles, never the internet. It’s also where you start capturing something worth keeping.

A library of proven thinking tools, such as a premortem, first-principles review or inversion. When something in your numbers moves, Struktürs suggests the one that fits. Your team decides whether to use it.

Three standards, earned rather than assigned. Working Analysis is there from day one: advice, with sources, that never blocks you. Operational Intelligence opens after three signed-off periods scoring 65 or more on average, and brings Decision Frameworks. Board-Grade opens after six signed-off periods scoring 72 or more: it won’t draft what it can’t source. You can see how close you are, and the whole team moves up together.

The progression reflects the maturity of your organisation’s financial record, not one person’s access level. As verification scores improve, Decision Frameworks open up for everyone; as knowledge articles and operational context accumulate, the reasoning sharpens for everyone who asks. It starts where you are and accelerates as more context is contributed.

Day one, a 30-minute call to check we’re a fit. By the end of the next business day after payment, your own Google Cloud Project is set up in your country. Then a 90-minute session with one of our team, where you upload your first trial balance and leave with it checked.

That’s where most teams start. Your first score will be low, and that’s useful, not a failure. Each month shows whether it’s improving, and Struktürs points to what to fix.

No. Free board seats (fair use) on Growth and Scale; on Enterprise by contract. A non-executive director isn’t a finance seat. Invite the whole board.

The verified layer only: content that has been admitted to the vault. Board packs, the figures behind them, and the reasoning that produced them. They never see drafts, a half-mapped trial balance, the review queue, or the Admin Console. You control what is finished; they see nothing else.

Because a board pack is only worth having once the numbers under it are solid. Signed-off financials, working papers and your team’s reasoning come first. Board Packs come with Growth and above; everything else is on every plan.

The reviewed analyst workpaper: every balance-sheet movement explained, with risks, notes and supporting papers attached. It is the record your auditor expects to see, and it is the bridge between raw financials and a board narrative. Most tools stop at a dashboard.

The reasoning your team would otherwise lose: retrospectives, risks, plans and decisions, captured in a conversation rather than a form. The AI asks the questions, drafts the article and links it to the figures behind it. Nothing is filed half-finished.

Not yet. Today you upload a CSV or Excel export from Xero, QuickBooks, Sage or any ledger, so there’s no ledger access to grant and nothing for IT to approve. Direct connections are on the roadmap.

First verified trial balance in an afternoon: upload, review the mapping, sign off. Working papers the same week. The compounding takes longer and is the point: mapping corrections carry forward, so period two is faster than period one. Board Packs are on Growth and up, and a board-grade pack needs six signed-off periods behind it, because a board paper drafted from two months of data wouldn’t survive the first question.

We’re working with a small first group of finance teams, deliberately. We’d rather work closely with a few than show a wall of logos. If you need a long list of references, we’re not the right choice yet. If you want a say in what gets built, now’s the time.

It can’t put anything on the record by itself. Every mapping, article, sign-off and board pack section needs a named person on your team to confirm it. If it can’t find a source, it tells you. Every draft shows its sources so you can check the reasoning, and the audit trail records who confirmed what, and when.

BI is for questions about numbers you already trust. This is about getting to numbers worth trusting, with the reasoning kept alongside. You could build the data pipeline yourself, but then you’d be running a data pipeline instead of running finance.

A plan for your team, with seats and a monthly allowance of AI work included. Seats are who can log in; tokens measure the AI work they ask for. Full plans are on the pricing page. See the plans.

You’re warned at 85% of your monthly allowance. At 100%, new AI work pauses until next month, or you can move up a plan. Nothing else stops, and your data is never at risk.

Yes. An Owner or Admin can change plan from the Admin Console. Upgrades apply straight away, and downgrades at the next renewal. Every plan starts with a 3-month pilot.

Your files, figures and database are stored in your country. You choose the country at sign-up: the United Kingdom (London) or South Africa (Johannesburg) today. Some processing happens elsewhere, such as the AI models, and the Security page lists each leg, where it runs and on what basis. Read the Security page.

Every customer gets its own Google Cloud Project, with its own database and file storage, in its own country. It isn’t a partition of a shared database. Storage is encrypted under a key held for your country. It’s one key for the country, shared by the companies there.

Yes. On request, we export your records and audit trail as open files (JSONL and Parquet), with a SHA-256 manifest of every file. Read Terms of Service §16.

ISO 27001-aligned controls from the start, with each company’s data stored in its own country from day one. We’re not yet certified, and we say so.

Each client is its own account, in its own country, so their data never mixes. What your team learns about each client’s close stays with that client, and the way you work comes with you from one to the next. Talk to us about how to set up several clients from the start.

The CFO or finance lead who wears several hats, and the team around them. You don’t need a team of analysts to start, just a trial balance and time to review the results each month. Each account is one company; groups with several use Enterprise.

Yorkville Advisors B.V. Every signal investigation and framework is written and reviewed by our finance experts before it ships.