fileAI
fileAI is an AI-native data preparation platform that turns unstructured enterprise files into validated, governed and traceable data, so finance, insurance and operations teams can run workflows and AI agents on records backed by citations and timestamps.
What is fileAI?
fileAI is the commercial brand of Bluesheets Pte. Ltd., a Singapore company, with a US entity, fileAI LLC (Delaware), in an administrative role. The platform positions itself as a governed execution layer between enterprise files and the AI agents meant to act on them: rather than stopping at extraction, it aims to deliver data that carries its own evidence.
The foundation product is fileForge, built on a four-stage chain: Capture, Prepare, Govern, Orchestrate. Capture combines zero-shot document classification with proprietary multimodal OCR and outputs structured markdown. Prepare runs an AI Schema engine that maps those outputs onto prompt-driven structures, then normalizes and reconciles them. Govern applies reasoning models that check each step, human-in-the-loop controls and in-context checklists. Orchestrate runs SOP-driven workflows and AI Query that route, match and execute.
Three proprietary capabilities are named. Scout maps a document and selects the sections worth processing. Forensics tests integrity, looking for tampering, anomalies and synthetic content. Resolve normalizes, enriches and matches records across sources. Querying happens through FQL (File Query Language) in natural language, which the publisher presents as requiring no code. Every extraction, validation and transformation is logged with citations and timestamps.
Two vertical products sit on top. fileLedger automates financial operations: accounts payable and receivable, multi-way reconciliation, close, ERP posting and a SuperAgent. fileShield covers insurance case management for insurers and TPAs, from claims and underwriting to contract administration and contract translation.
Accepted formats include JPEG, PNG, TIF, TIFF, HEIC, HEIF, PDF, CSV, XLSX, XLS, DOC, DOCX, ZIP and WEBP. The publisher claims more than one billion files processed, over 200 languages plus handwriting, 500+ large enterprise customers, 100+ ERP and system integrations, and up to 90% of the time saved on workflow execution. Those figures are the vendor's own and are not independently verified.
Two scope limits are worth stating. Third-party model providers, including OpenAI, Google, Anthropic and Perplexity, are engaged as subprocessors, so content passes through them under contract. And the product is web and API only: public documentation covers a User Guide, the API and an MCP server, with a Discord community, but there is no mobile app.
What it does
- Map and classify any incoming file automatically before extraction, with Scout selecting the sections worth reading.
- Extract structured data into clean markdown, with citations and timestamps attached to each output.
- Detect tampering, anomalies and synthetic content at ingestion with Forensics.
- Normalize, enrich and reconcile records across files and systems with Resolve.
- Query documents in plain language through AI Query and FQL, with no code required.
- Orchestrate SOP-driven workflows with human review on risky or low-confidence cases.
- Publish validated outputs to ERPs, AI agents and downstream decision systems.
When to use fileAI / When not to
A quick filter to help you decide if fileAI is the right fit.
When to use fileAI
- Finance and accounting teams automating accounts payable and receivable, multi-way reconciliation, month-end close and ERP posting.
- Insurers and third-party administrators processing claims, underwriting submissions, policy administration and complex case files.
- Banking, compliance and risk functions handling KYC packs, credit files, trade finance documents and other regulated paperwork.
- Healthcare, laboratory and life-sciences organizations extracting data from clinical documents, EMR exports and regulatory records.
- Data and AI teams, along with supply chain, procurement and BPO back offices, that need governed, citation-backed data to feed agents and downstream systems.
When not to use fileAI
- Individuals and small teams looking for a free, unlimited document reader: the Self-serve plan is billed per page processed, and no per-page rate is published.
- Anyone who needs a mobile app or a browser extension: fileAI ships as a web application and an API only.
- Teams that need a localized product: the site, the interface, the documentation and the support channels are English-only.
- Buyers who require custom-trained OCR models, on-premise deployment or a 99.9% uptime SLA at a published price, since all of that sits in the quote-only Enterprise plan.
- Anyone wanting to benchmark, resell or study the platform competitively: the acceptable use policy forbids reverse engineering and competitive use.
How to use fileAI
A typical end-to-end flow, from setup to results.
- Pick your entry point: create a Self-serve account online, or request a guided demo if you need Enterprise scoping.
- Create a workspace by following the documented quickstart.
- Choose how files arrive: email, direct upload or the API, with unlimited import integrations on the Self-serve plan.
- Define an AI Schema describing the fields you want, written as a prompt rather than mapped by hand (up to two schemas on Self-serve, unlimited on Enterprise).
- Let Scout map the incoming files and select the sections worth extracting.
- Review the validation checklists and the citations attached to each field, and clear exceptions through human-in-the-loop review.
- Interrogate the processed documents in natural language with AI Query and FQL before you automate anything.
- Connect an export integration to your ERP or downstream system (one export on Self-serve, bespoke on Enterprise).
- Automate the flow with SOP-driven workflows, approval rules, routing and escalation.
- Lean on the public documentation (User Guide, API, MCP server), support by email and the Discord community; discuss private cloud, on-premise deployment or custom-trained OCR models with the vendor when needed.
Pros & Cons
Pros
- Traceability is native: citations, timestamps and an audit trail on every extraction, validation and transformation.
- Document integrity checks for tampering, anomalies and synthetic content, which few tools in this segment offer.
- Unusually open legal posture: a complete public DPA incorporated into the terms, and a named subprocessor list with addresses and locations.
- Ownership of Customer Data is recognized explicitly in the contract, with fileAI's license limited to delivering the service.
- Broad claimed compliance coverage: ISO 27001, SOC 1 Type II, SOC 2 Type II, HIPAA and GDPR alignment.
- Multi-model orchestration to arbitrate between quality, speed and cost, with private cloud and on-premise deployment options.
- A $0 per month Self-serve entry point to test, and named, quantified customer references including MSIG, Forvis Mazars, Nippon Paint, Little Farms, Koala and Osome.
Cons
- The Self-serve plan is billed per page but the amount charged per page is never published, so the real cost cannot be estimated before contact.
- Enterprise and fileShield pricing is quote-only, and the fileLedger figures are shown as "from" prices.
- The certifications on display can only be checked through a Vanta-hosted Trust Center that renders in JavaScript and could not be read, so they remain unverified claims.
- No EU representative is designated under Article 27 of the GDPR, and primary storage defaults to Singapore, with EU data residency available only as a paid option on request.
- No dedicated contact page: every inbound route goes through the demo form.
- Web and API only, with no mobile app or browser extension, and a site, interface and documentation available in English alone.
- Free trials are left to the publisher's discretion under clause 5.6 and are not advertised on the pricing page, while fees paid are non-refundable except after a material change to the terms (clause 9.4).
Pricing & Plans
There is a free entry point. The fileAI platform's Self-serve plan is listed at $0 per month, with usage billed per page processed, one billable page being one document page or one spreadsheet sheet; the site does not publish the per-page rate, so the effective cost cannot be estimated in advance. The lowest published paid price is fileLedger Lite, from $11 per month for 2,000 pages, followed by fileLedger Standard from $21 per month for 5,000 pages. fileLedger Business, the Enterprise platform plan and fileShield are quoted on request. Fees are exclusive of taxes and invoiced in advance for each period, subscriptions renew automatically with 30 days' notice required to change or cancel, late payment carries interest of 1.5% per month with possible suspension beyond 15 days, and fees paid are non-refundable except where the terms are materially amended.
- several vLM AI OCR models
- up to two AI schemas
- all supported formats
- handwriting and 200+ languages
- ingestion by email
- API or upload
- unlimited import integrations
- one export integration
- everything in Self-serve plus custom-trained OCR models
- unlimited schemas
- bespoke exports
- customizable orchestration
- private cloud or on-premise deployment
- a 99.9% uptime SLA and a dedicated account manager.
- unlimited users
- Xero integration and CSV export
- SSO
- role-based access control and audit trail.
- Xero and QuickBooks integrations
- plus assisted onboarding.
- Xero
- QuickBooks
- NetSuite and further integrations
- files over 100 MB and a dedicated account manager.
- fileShield (insurance case management)
- price on request. A monthly/annual toggle is offered on the fileLedger grid.
Data, GDPR & hosting
A consolidated view of how fileAI handles your data.
GDPR overview
GDPR is addressed explicitly and in detail. Bluesheets Pte. Ltd. (138 Robinson Road, #26-01 Oxley Tower, Singapore 068906, UEN 201605699K) is named as controller, with a data protection officer reachable at privacy@file.ai or dpo@file.ai. The privacy policy (version 2.0, May 2026), the legal terms (version 4.0, April 2026) and the public DPA incorporated by reference cover transfers outside the EEA through the European Commission's standard contractual clauses, the ICO UK Addendum and adequacy decisions; the SCCs are governed by Irish law and Irish courts. Access, rectification, erasure, restriction, portability, objection, withdrawal of consent and complaint to a supervisory authority are listed, with a one-month response target (45 days in California). Singapore's PDPA and California's CCPA/CPRA have dedicated sections, the subprocessor list is public (version 1.0, 1 January 2026), and breach notification to users and regulators is provided for. One gap: no EU representative under Article 27 is designated.
Who owns the data?
Under the legal terms, you keep every right in your Customer Data. fileAI receives only a limited license to process, host, store and use that data to deliver the service and perform its obligations, and cannot use it for any other purpose without your consent (clause 3.1). Clause 3.3 reserves to fileAI and its licensors all rights in the service itself, including its models, algorithms and training data. The roles split accordingly: Bluesheets Pte. Ltd. acts as controller for account data, and as processor for the personal data contained in the files you upload, under the publicly available data processing agreement incorporated into the terms by reference.
Reuse rights
The privacy policy (version 2.0, last updated May 2026) lists the purposes: creating and managing accounts, providing and improving the service, billing, communications, security and fraud prevention, analytics and legal compliance, on the legal bases of contract performance, legitimate interest, legal obligation and consent. On training, the line is drawn explicitly: Usage Data and Aggregated Data, which cannot identify you, may be used to improve and train fileAI's AI models, while your personal account data and the content of the documents you upload are not used for training without your explicit consent. Clause 3.4 of the terms allows Aggregated Data and Usage Data to be used to operate, maintain and improve the service, and the contractual definition of Aggregated Data expressly excludes Customer Data. fileAI states that it does not sell personal data and does not share it in the CPRA sense, and that no automated decision producing legal effects is taken without notification. As the customer, you keep your data and can export and reuse it without asking permission, within the limits of the acceptable use policy.
Data retention & training
Hosting summary
Primary storage defaults to AWS Asia Pacific (Singapore), under clause 10.7 of the DPA, with backups distributed geographically across the fileAI regions available. Regional hosting and data residency options exist but are chargeable and granted on request through sales@file.ai, which means EU residency is never the default. The privacy policy states that personal data is processed mainly in Singapore and the United States via AWS. The published subprocessor list adds Australia and the Netherlands to that footprint: AWS, OpenAI, Google, Anthropic, LangChain, ClickHouse and Perplexity in the United States, Singapore and Australia, and Posthuman CMF in the Netherlands. Access is possible by fileAI staff in Singapore and in the other countries where the company operates. Technically, data is encrypted in transit with TLS and at rest with AES-256, hosted on ISO 27001 certified infrastructure with redundancy across several availability zones. Private cloud and on-premise deployments are offered for customers who need full isolation. The exact boundary between where data is stored and where it can be accessed is not fully documented.
Where fileAI works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting fileAI.
- The Self-serve per-page rate is not published, so the real bill depends entirely on volume; the fileLedger prices are "from" figures and the monthly/annual toggle never spells out the annual amount.
- Subscriptions renew automatically and require 30 days' notice to stop or change, and fees already paid are non-refundable (clauses 9.1 and 9.4).
- A free trial may be granted at the publisher's discretion and converts automatically into a paid subscription if it is not cancelled in time (clause 5.6).
- US customers are bound to mandatory AAA arbitration in Wilmington, Delaware, and waive class actions, with a 30-day written opt-out to legal@file.ai.
- Data residency in the EU is possible but paid and on request, the default being Singapore, and no EU representative is designated under Article 27 of the GDPR.
- The certifications on display are only accessible through a JavaScript-rendered Vanta Trust Center, and the UEN published on the site (201605699K) differs from the one held by third-party registries (201925741W).
- AI outputs must be checked before use (clause 7.4): citations and audit trails make review possible but do not replace it, and the more reliable an automated pipeline feels, the easier it becomes to stop looking at what it produces.
Setup & Integrations
Technical difficulty
Easy to start, substantial to industrialize. Self-serve sign-up is immediate and the quickstart runs in minutes: files can arrive by email or direct upload with no integration work, AI Schemas are written as prompts rather than mapped by hand, FQL queries need no code, and extraction is advertised as requiring no prior training. The real effort appears later, in SOP-driven workflows, approval rules and ERP integration, which are project work rather than configuration. Assisted onboarding starts with fileLedger Standard and a dedicated account manager comes with Business and Enterprise; private cloud or on-premise deployment is a full infrastructure project.
Deployment
Integrations
Supported languages
Behind fileAI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Which file formats does fileAI accept?
Does it handle handwriting and non-English documents?
Can it process long documents?
How does per-page billing work?
How is data secured?
Where is the data stored?
Are my documents used to train the models?
Is there an API?
Is there a DPA and a list of subprocessors?
Is there a minimum age, and is there a mobile app?
Should you pick fileAI?
fileAI is enterprise software, not a consumer utility, and it reads that way from the first page. What it really sells is less extraction than governance: Scout, Forensics and Resolve, the validation checklists, the citations and the timestamps all exist so that a finance, claims or compliance team can show why a figure ended up in the ERP. If your problem is that nobody can audit how a number got there, that is the strongest argument on offer.
The publisher's legal posture is more open than most: a full public DPA, eight named subprocessors with addresses, explicit contractual recognition that Customer Data stays yours, and named, quantified customer references such as MSIG, Forvis Mazars, Nippon Paint, Little Farms, Koala and Osome. Bluesheets Pte. Ltd. operates from Singapore with a New York presence, and third-party sources put total funding somewhere between USD 20 and 26 million.
The weak point is commercial transparency. Past the $0 Self-serve entry point, pricing thins out fast: the per-page rate is never published, fileLedger prices are "from" figures, and Enterprise and fileShield are quote-only. The certifications on display, ISO 27001, SOC 1 and SOC 2 Type II and HIPAA, sit behind a Trust Center that renders in JavaScript and could not be read, so they remain claims rather than verified facts. Two further details matter to European buyers: no Article 27 EU representative is designated, and EU data residency is a paid option layered on top of default Singapore storage.
Credible, then, for regulated and document-heavy operations that need an audit trail as much as an extractor, provided you get pricing in writing, confirm the certifications directly and settle data residency before signing.
- Choosing a selection results in a full page refresh.
- Opens in a new window.