FYI and the Document Substrate: What Each Does and Where They Overlap

Many accounting practices in New Zealand and Australia run on FYI, and a common question from a principal is how the Document Substrate differs from it and whether the two would compete. They do not. FYI files the documents and runs the practice. The Document Substrate reads what was filed, checks it, and gives a reviewer a workpaper with every figure traced to its page. This article sets out where the two overlap, where they differ, and what a practice on FYI gets from using both.


What FYI does

FYI is a cloud document and practice management platform for accounting firms in Australia, New Zealand and the United Kingdom. Its modules cover email and document filing, search, client records, tasks and comments, job management, document review and approval, client sharing through a portal, forms, process automation, time and billing, reporting and employee management. Client and job records sync from Xero Practice Manager, or from MYOB, APS, CCH or GreatSoft on the lower plans.

The feature the platform is built around is AutoFile. Email in every user's Outlook inbox and sent items is filed against the matching client, roughly every 15 minutes, by rules the practice sets. Documents arrive by upload, through the Outlook and OneDrive add-ins, by client upload through the portal, from signing services and the ATO, and by migration from whatever the practice used before.

In my observation, most practices use FYI for a smaller set of things than its feature list: AutoFile, document storage and search, client and job records, tasks and comments, and on the top plan, time and billing. Forms, the client portal, custom automations, document approval workflows and FYI AI are used by a minority.

FYI AI is in beta and available on the Elite plan only. It is not a chat. It is a set of steps inside a custom automation: a prompt over the triggering document or email, a drafted email whose body comes from that prompt, a set of tasks created from one prompt, and automatic filing using the cabinet, keywords and categories the prompt returned. The marketed uses are draft email replies, task delegation, automatic filing, document summaries and meeting summaries.

What the Document Substrate does

The Document Substrate is the software between the documents a client sends and the ledger. It applies the firm's captured expertise, its substrate, to those documents. It does one job, in six steps.

  • Reads the pages. PDF, Word, Excel, scans and photos. A page-reading service produces accurate text and layout, including tables, before anything else happens.
  • Removes personal details. Names, addresses, bank account, IRD and company numbers are replaced with placeholders before the model that answers sees the text. The page-reading service and the redaction model see the original, because their job is to read it. Real values are kept in an encrypted vault and restored only for people allowed to see that client's documents.
  • Reads the figures. Twenty-two document types, from bank statements to rental summaries, each with a defined set of fields. Every figure is traced to a page and the exact characters it came from. A figure that cannot be found on the page is marked and left for a person, not passed silently.
  • Checks them by code. Forty-nine checks on single documents run in code, not in the model: a statement must foot from opening to closing balance, an invoice's lines must add to its total, dates must fall inside the period. Rules across documents check one document against another.
  • Shows findings to a reviewer. Every check that fires, and every figure the machine was unsure of, goes into one queue with a plain-language reason. A person accepts, corrects or resolves each one and signs the job.
  • Produces the output. A workpaper, a draft reporting letter, Excel and Word exports, and a chase list of what the client has not yet sent, all generated from checked figures only. Questions over the documents are answered with citations that open the page.

It is not a filing system and it is not practice management.

How the two fit together

FYI is the system of record and the Document Substrate works beside it. FYI keeps filing. The Document Substrate reads from FYI's storage, on a schedule, and returns the checked work.

FYI files and runs the practice on the left; the Document Substrate reads, checks and returns work on the right; the connector between them is configured for each practice

The connector is configured for each practice. FYI's external API lets an application list, fetch and download documents, read client and job records, create new documents, and write values onto a job's custom fields. The connector uses that API with access keys the practice issues from its own FYI account, so setting it up is part of implementation. A practice without FYI sends documents by upload or to a per-user email address instead.

Documents are picked up on a schedule. FYI can send a notification the moment a document is filed, but only as an automation step, which needs the Pro or Elite plan and someone at the practice to configure it. The connector therefore works by asking FYI what has changed since the last check. That works on every plan.

The return path matters most. Without it, a preparer downloads from one system and uploads into the other, and that manual step is what the connector is there to remove. Three things go back: the reporting letter filed into FYI as a new document, the job's check status written onto the FYI job's custom fields, and the chase list delivered as an FYI email draft or task. FYI's own filing is never modified. The connector only creates new records in FYI.

Where they overlap

The overlap is small. Five capabilities appear on both sides, and in each case the Document Substrate keeps the minimum its own job needs and does not extend it, because for an FYI practice the filing already happened in FYI.

  • Filing structure. FYI has cabinets, categories, year, client and job. The Document Substrate has collections and tags. A cabinet, client and job map onto one collection at import. No hierarchy and no custom fields will be added.
  • Summaries. FYI AI can summarise a document as an automation step on the Elite plan. The Document Substrate writes a summary for search and for the file drawer. It stays an internal step and is not marketed as a feature.
  • Automatic filing. FYI AI can file a document by the cabinet and keywords the model returned. The Document Substrate detects the document type and applies it when confident. That stays, because reading the figures depends on knowing the type. Filing metadata is never written back into FYI.
  • Asking the client for documents. FYI has the portal, forms, checklists, bulk email and reminders. The Document Substrate derives a chase list from what is missing, which FYI cannot do because it has not read the documents. The derivation stays. The sending happens in FYI.
  • Letters. FYI has templates, stationery, mail merge and PDF editing. The Document Substrate generates a letter from the checked figures, editable and aware of sign-off. The generation stays and the letter is filed into FYI. No editor, stationery or PDF tools will be built.

Two words mean different things in the two products. In FYI, a workflow is document review and approval status, and a job is a practice work item with states, budgets and time. In the Document Substrate, the same words mean the set of checks that run on a workpaper, and the workpaper itself. In material for FYI practices I use the checks and the workpaper, and the practice manual carries a glossary.

Where they differ

FYI, as its own documentation describes it, does not read figures out of documents into fields, trace a figure to a page, check arithmetic within a document, reconcile figures across documents, answer questions over a set of documents with citations, remove personal details before a model sees them, or compile a workpaper. Its completeness checking is a checklist a person or an automation sends to the client. Its AI features summarise, draft and file. They do not verify.

The Document Substrate does not file email, hold client records, manage tasks, track time, bill, run a client portal, sign documents, edit PDFs or report on the practice. For an FYI practice, all of that is FYI's job. For a practice without FYI, it is a practice management need the Document Substrate does not cover.

Neither product does the other's main job.

How the AI differs

Both products use Anthropic's Claude models with no data retention. The difference is not the model.

Two paths: FYI AI sends the document to the model whole and produces a draft; the Document Substrate redacts first, the model sees placeholders, code checks the figures, and a person signs

FYI AI sends the document or email to the model as it is, names and account numbers included, and the output is a summary, a draft or a filing decision that nobody checks against the source. That is a reasonable design for drafting a reply.

The Document Substrate sends the model redacted text only. The one exception is the page-reading service, which has to see the original file to produce the text. After the model has read the figures, code checks them, and a person resolves what code could not.

The difference is two steps: redaction before the model, and verification after it. Those two steps are what let the output go into a workpaper a partner signs.

Both products need the practice's sign-off on where data is processed. FYI processes in an Australian or United Kingdom region. The Document Substrate runs in accounts the practice owns, with the page-reading and model services named and their retention terms confirmed as part of set-up.

What this means for a practice on FYI

Nothing changes in FYI. Staff keep filing, keep working from the client workspace, keep their tasks and time where they are.

Reviewers get one more place to look. A queue of findings, each with a reason, and a workpaper per job. Each document in it carries its FYI reference and a link back.

The checked work comes back to FYI. The letter as a document, the status on the job, and the chase list as a draft or task.

Personal details stay out of the model. For a practice considering FYI AI, that is the difference to cite when a client asks about privacy.

Email and upload remain available. For a one-off document, or for a practice that does not run FYI, files can still be sent to a per-user email address or uploaded directly.

What Orbital will not build

Each item is FYI's job.

  • Client portal or forms. FYI has both. A portal sends personal details outside the practice by design.
  • E-signing. FYI integrates FuseSign, Annature and Adobe Sign.
  • Tasks, comments, time, billing, invoicing, practice reporting, capacity planning. FYI's core and its Elite modules. None of them involve reading documents.
  • An Outlook add-in or an AutoFile equivalent. AutoFile is the reason practices buy FYI.
  • PDF editing, templates, stationery, mail merge. The letter is generated from checked figures. Its letterhead is a setting, not a template system.
  • Client records. FYI's client list, synced from Xero Practice Manager, is the record.
  • AI email drafting, meeting summaries, AI task creation. FYI AI's feature set. None of it verifies anything.
  • Posting to the ledger. The Excel export is where the Document Substrate's work ends. The reviewer decides what goes to Xero.

The Document Substrate in detail · Principles · Explained simply · Project detail