
FabricAI
FabricAI is a self-learning purchase invoice automation layer for Finnish accounting and property management firms. It codes, VAT-classifies and routes supplier invoices inside existing accounting software, and claims up to 90% of invoices posted with no human review.
What is FabricAI?
FabricAI is a machine learning system that automates purchase invoice processing, published by Visma Solutions Oy in Finland. It presents itself as the first purchase invoice processing and automation system that learns by itself, with no rule libraries to build and no continuous maintenance to fund.
The product does not replace your accounting software, it sits between the arriving invoice and the ledger. It connects to Procountor, Netvisor, Fennoa, Fivaldi and Hausvise, reads incoming purchase invoices and predicts the coding line by line: accounts, VAT treatment, accruals, dates, cost centres and dimensions. It then applies the firm's rules, dispatches the invoice into the approval workflow and writes the finished entry back. A single browser portal gathers the invoices of every client an accounting firm serves into one shared view, which the vendor presents as a way to smooth holiday cover and level quality between files.
Progress is framed as a ladder. First the automation rate climbs. Then comes No Touch, where a person still sees and approves the invoice without touching it. Then No Look, where the invoice is handled from start to finish without a single human seeing it, and FabricAI places itself explicitly in the Human Out of the Loop category rather than human in or on the loop. The stated ambition is 90% by 2030.
The numbers are the vendor's own claims. The home page cites 1.5 million invoices a month and 60,000 companies; the English page claims more than 24,000 Nordic companies and above 95% accuracy on over a million invoices in 2022; the Lite page states 92.85% processing accuracy; a published sample of 40 companies over eight months gives 42.6%. They do not measure the same quantity over the same period. Under the hood, the FAQ describes an AI built on TensorFlow and trained on Nvidia V100 GPUs, and the home page says it accepts instructions in plain Finnish.
The primary targets are accounting firms and property management companies, with named references spanning large practices, companies and municipalities, and a claimed NPS of +70 measured in September 2023. A partner branded version exists under the Neverlook name for software vendors. Go-live is announced in a matter of hours, with roughly one hour of accountant training.
What it does
- Pull purchase invoices out of the accounting software through a restricted API connection
- Predict the full coding line by line: accounts, VAT, accruals, dates, cost centres and dimensions
- Process and post safe invoices end to end without any human looking at them, the No Look tier
- Send invoices into the approval workflow automatically, following rules defined per company
- Route risky invoices back to the accountant, such as large amounts or uncertain predictions
- Write the finished entry back into the accounting software
- Track automation in an analytics portal and push the figures to a BI tool or spreadsheet
When to use FabricAI / When not to
A quick filter to help you decide if FabricAI is the right fit.
When to use FabricAI
- Accounting firms of any size, from small practices to large groups, that process high and repetitive volumes of supplier invoices
- Property management companies handling purchase invoices across large portfolios of housing companies
- Bookkeepers and accounts payable teams already working inside Netvisor, Procountor, Fennoa, Fivaldi or Hausvise
- Finance departments in SMEs, larger companies and municipalities that want invoice coding handled without adding headcount
- Software vendors and integration teams that want to embed invoice automation through the AI Inside REST API
When not to use FabricAI
- Anyone shopping for an accounting system or ERP: FabricAI is designed as an intelligent automation layer, not as a primary system of record
- Teams that need sales invoicing, payroll or payment runs automated, since the API rights granted cover purchase invoices only
- Organisations outside Finland, where no pricing is published and the supported accounting systems are all Finnish
- Firms whose accounting software blocks API-driven dispatch into the approval workflow, because that step then cannot be automated
- Buyers who want to test before committing, or who need a mobile app: there is no free plan, no free trial and no iOS or Android application
How to use FabricAI
A typical end-to-end flow, from setup to results.
- Book the free 15 to 30 minute demo to see the automation applied to your own invoice flow
- Agree the commercial terms: a fixed monthly fee plus 0.38 EUR per No Look invoice, with guided onboarding at 690 EUR as an option; recurring billing starts once the service is live for you
- Create a dedicated API user in your accounting software and grant it read and write rights limited to purchase invoice data
- Attach your client companies one by one, since FabricAI only reaches the companies explicitly connected to it
- Let the AI train on your invoice history, the first of the three implementation stages
- Train the bookkeepers, which the vendor budgets at roughly one hour
- Open the browser portal with the email and password created at go-live, on an up-to-date Firefox or Chrome under Windows, Mac or Linux, and switch on two-factor authentication
- Set your automatic approval thresholds and define the approval workflow rules company by company
- Start the coding automation and keep correcting the proposals, since the model learns from every correction the accountants make
- Follow accuracy and automation rates in the analytics portal, export them to BI, or read them through the AI Inside REST API documented for developers
Pros & Cons
Pros
- Self-learning by design: no rule library to build and no rule maintenance to keep paying for
- Outcome-aligned pricing, since an invoice that still needs expert work is not billed, backed by a 100% satisfaction guarantee under which an unsatisfied customer does not pay
- Keeps the accounting software already in place, so no new tool has to be imposed on end clients
- Fast go-live measured in hours, with accountant training of about one hour
- Vendor-stated gains of 2 minutes down to 5 seconds and 0.75 EUR down to 0.25 EUR per invoice, and up to seven working days saved per month
- One multi-client view that makes stand-in cover easier and levels quality across client files, plus analytics and an open REST API to build further services on
- Servers and data in the European Economic Area with TLS encryption, a named and dated sub-processor list published in the terms, multilingual support by chat, phone and email, and the backing of Visma Solutions Oy
Cons
- No standalone privacy policy on the domain: the address linked from the FAQ and the cookie notice returns a 404
- The terms explicitly permit the processed data to be used to train the AI model, and no opt-out mechanism is documented anywhere
- Two sets of terms coexist on the site, the 2022 terms of use naming FabricAI Oy and the 2024 service terms naming Visma Solutions Oy, and two support addresses circulate with them, moro@fabricai.fi and tuki@fabricai.fi
- No free plan and no trial period: the fixed monthly fee falls due from go-live
- Pricing is barely public. The only published entry price, 180 EUR per month, sits on the Lite page reserved for Netvisor firms and capped at 500 invoices, while the main pricing page shows only a simulator alongside the 0.38 EUR per No Look invoice and the 690 EUR onboarding fee
- Site, pricing and integrations are built around the Finnish market, with no price list published outside Finland, and there is no mobile application
- The Visma terms exclude any guarantee of results and any refund of amounts already invoiced; objecting to a new sub-processor only earns the right to terminate within 30 days rather than the right to have it dropped; and retrieving or deleting data at the end can be billed at the publisher's hourly rate
Pricing & Plans
There is no free plan and no free trial. Pricing combines a fixed monthly fee covering the portal with per-invoice billing: 0.38 EUR for each invoice processed under No Look automation, while any invoice that still requires an expert's intervention is not charged. The lowest published entry point is FabricAI Lite, from 180 EUR per month plus 0.38 EUR per No Look invoice for up to 500 purchase invoices, reserved for accounting firms using Netvisor. Guided onboarding and training are an optional 690 EUR. Above 5,000 invoices a month, pricing moves to an Enterprise quote. Finnish VAT is added, the month elapsed is invoiced on the first working day of the following month, and payment terms are 14 days net. Price increases carry two months' notice under the Visma service terms, one month under the 2022 terms, with a right to terminate in either case. A 15 to 30 minute demonstration is free and without commitment, and the vendor advertises a 100% satisfaction guarantee.
- from 180 EUR per month plus 0.38 EUR per No Look invoice
- for up to 500 purchase invoices
- reserved for accounting firms using Netvisor
- with go-live announced in a single day
- a fixed monthly fee calculated by the on-site simulator from your monthly volume and expected No Look rate
- plus 0.38 EUR per No Look invoice
- up to 5
- 000 invoices a month
- with guided onboarding available at 690 EUR
- quoted individually above 5
- 000 invoices a month
- priced on volume
- number of sites and expected No Look rate
- with discounts depending on the level of support required and the spread of the client portfolio
- the portal can also be used without No Look automation
Data, GDPR & hosting
A consolidated view of how FabricAI handles your data.
GDPR overview
GDPR compliance here is contractual rather than promotional. A data protection annex is built into the Visma service terms and cites EU regulation 2016/679 explicitly. Clause 7.1 commits to organisational, technical and physical safeguards matching Article 32; clause 4.5 promises assistance with data subject requests and with Articles 32 to 36; clause 4.9 covers breach notification without undue delay. Sub-processors are named in Annex B under a general authorisation, with 30 days to object to a new one and standard contractual clauses for transfers outside the EU or EEA. Clause 8.1 grants one audit per year at the customer's cost, answerable with an ISAE or ISO report under 12 months old. The gap is visible: no standalone privacy policy is published on fabricai.fi, and the address cited by the FAQ and the cookie notice returns a 404. Established in Finland, the publisher needs no Article 27 representative.
Who owns the data?
Under the data protection annex built into the Visma service terms, the customer is the controller of everything it stores in the service and Visma Solutions Oy acts only as processor. The customer warrants the accuracy, integrity, content, reliability and lawfulness of the data it sends. Purchase invoices are declared portable customer data under the EU Data Act, so they belong to the customer. Machine learning models, algorithms, system parameters and security logs stay with the publisher, covered by its intellectual property rights and explicitly non-portable. One caveat sits in clause 4.13: for aggregated usage data, Visma declares itself controller, moving that data outside the processing agreement.
Reuse rights
The customer can reuse its own material freely and without asking permission. Processed invoices are written straight back into the accounting software, and purchase invoice data can be exported in JSON or CSV on request to support. The other direction deserves a careful read. Annex A, section A.5 of the service terms lists the use of data to train the AI model among the processing operations performed, and clause 4.14 has the controller accept, and not object to, the use of customer data in combined and anonymised form for improving the service, for research, for training and for statistical purposes. Clause 4.13 covers aggregated usage data used to optimise the service and build new features. Learning also happens customer by customer: the model picks up the accountant's corrections to refine rules specific to each company. On the sharing side, the FAQ states that no customer data, whether personal, accounting or invoice data, is disclosed to parties outside the company or to stakeholders including investors, and access to the accounting software runs through a restricted API user limited to purchase invoice data, company by company.
Data retention & training
Hosting summary
The FAQ states that the service's servers and the data collected through the service are located in the European Economic Area, with transfers encrypted in TLS; local media are encrypted and kept in locked premises accessible only to authorised staff. No individual hosting country is named. Annex B of the service terms, in its 17 September 2025 version, names the sub-processors: Amazon Web Services (EU/EEA, server storage), Freshworks Inc (EU/EEA, customer service system), Google Gemini (EU/EEA, AI features), MailerSend Inc (EU/EEA, email service) and Microsoft (EU/EEA, authentication services and Microsoft Azure), plus Google Firebase (United States, authentication services) covered by standard contractual clauses. The current list is maintained on the Visma Trust Centre. Two clauses shape the real perimeter: group companies located in the EU or EEA may act as sub-processors without appearing on the list and without prior notice, and any transfer outside the EU or EEA runs on the Commission's standard contractual clauses. The publisher states that it verifies the security level of third-party computing and storage services before transferring data to them.
Things to keep in mind
Risks and trade-offs to weigh before adopting FabricAI.
- Your data trains the model and there is no way out. Annex A, section A.5 of the service terms lists AI model training among the processing operations, and clause 4.14 has you accept, and not object to, aggregated and anonymised use of your data for improvement, research, training and statistics. No opt-out mechanism is documented anywhere on the site or in the terms.
- No privacy policy is published on the domain. The link cited by the FAQ and by the cookie notice returns a 404, so the only data protection documents you can actually read are the cookie notice and the annex buried inside the service terms.
- The site contradicts itself on who you are contracting with. The 2022 terms of use name FabricAI Oy as the provider, the 2024 service terms linked in the footer name Visma Solutions Oy, and the Finnish trade register confirms that FabricAI Oy ceased to exist as a company name on 1 September 2024. Two support addresses circulate in parallel, moro@fabricai.fi and tuki@fabricai.fi. Establish which document binds you before signing.
- The service commitments moved backwards between the two documents. The 2022 terms promise a 99% SLA and a two-hour response; the 2024 Visma terms exclude any availability guarantee and provide the software as is, with no guarantee of results and no refund of amounts already invoiced.
- The automation percentages are not comparable with one another. Up to 90%, 92.85% and above 95% come from different pages, different periods and different quantities, while the one detailed sample published, 40 companies over eight months, gives 42.6%. Modelling your business case on the headline number is the easiest way to be disappointed.
- Human out of the loop is the explicit goal, and that has a human cost. When No Look means nobody ever sees the invoice, the skill and the reflex of reading one fade, and over-trust settles in quietly. If the model drifts, or an atypical supplier slips through, there is no one watching. Keep approval thresholds tight, keep sampling invoices by hand, and keep reading the analytics portal.
- Exit and continuity clauses bite. You have 30 days after the contract ends to request your data back, retrieval and deletion can be billed at hourly rates, objecting to a new sub-processor only buys the right to terminate within 30 days, the publisher reserves the right to withdraw software from a market on 12 months' notice and to use your name and logo as a reference, and Google Firebase authentication sits in the United States under standard contractual clauses.
Setup & Integrations
Technical difficulty
Low for the accounting firm. Nothing is installed: FabricAI is a browser application running on an up-to-date Firefox or Chrome under Windows, Mac or Linux. The only technical step on your side is creating a dedicated API user in your accounting software and granting it rights restricted to purchase invoice data. Implementation runs in three stages, training the AI, training the bookkeepers, then starting the automation, with go-live announced in hours and about an hour of training; guided onboarding costs 690 EUR. The developer route is more demanding: integration through the AI Inside REST API.
Deployment
Integrations
Supported languages
Behind FabricAI
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Is there really an AI behind FabricAI?
How is FabricAI set up in an accounting firm?
Which accounting systems does it work with?
What does FabricAI cost?
Is there a free trial?
Is my data used to train the AI model?
Where is my data hosted?
Can I get my data back if I leave?
How much automation should I actually expect?
Who publishes FabricAI?
Should you pick FabricAI?
FabricAI does one thing and pushes it to its logical end. It takes the purchase invoice pile away from the bookkeeper, and in its No Look tier it removes the human gesture entirely. That focus is both the strength and the boundary: it is not a ledger, not a payments system, and it deliberately stays an automation layer sitting on top of the accounting software you already run.
The traction claimed is unusually concrete for this category: 1.5 million invoices a month, 60,000 companies, more than 10% of Finnish accountants as monthly users and an NPS of +70 measured in September 2023. None of it is independently audited, and the automation figures quoted across the site, from up to 90% down to 42.6% on the only detailed sample, do not measure the same thing.
The commercial model is genuinely buyer-friendly. You pay a fixed monthly fee plus 0.38 EUR for every invoice the AI closes on its own, an invoice that lands on an accountant's desk costs nothing, and the vendor adds a 100% satisfaction guarantee. Since 1 September 2024 the product has sat inside Visma Solutions Oy, which buys durability at the price of standard group terms that are unlikely to be negotiable.
Two reservations deserve a decision before signing. First, the service terms list AI model training among the processing operations and have the customer accept aggregated, anonymised use of its data, with no documented opt-out. Second, no standalone privacy policy is published on the domain at all. Add the de facto Finnish scope, in integrations, pricing and sales language, and the audience is clear: a Finnish accounting or property management firm drowning in repetitive supplier invoices, and comfortable with its data feeding the model.
- Choosing a selection results in a full page refresh.
- Opens in a new window.