Stop chasing invoices: how Tripletex/Fiken invoice processing can be automated with AI

AI-based invoice automation is now available directly in Tripletex and Fiken — not as future functionality, but as active integrations that Norwegian accounting firms are already using. The systems read incoming invoices, propose bookings, and send approval requests without manual data entry. Firms that have implemented such setups free up many hours per week per accountant — time spent instead on advisory work and client contact. The technology is mature, the costs are manageable, and the integration work is far simpler than it was just a few years ago.

Invoices eat time you don't have

Invoice processing is among the most time-consuming routine tasks for accounting firms and SMBs with high transaction volumes. The process is simple to describe — and that simplicity is precisely what makes it vulnerable.

A typical accounts payable cycle looks like this:

1. Invoice arrives (email, EHF format, postal mail, or via vendor portal)
2. Someone reads in vendor data, amount, VAT, and due date
3. Invoice is posted to the correct account
4. Approver reviews and approves
5. Invoice is paid on time
6. Document is archived and accounts are closed

Each step is a potential bottleneck: an invoice landing in the wrong inbox, an approver on holiday, an incorrect posting. One error can cost many hours of correction work and result in delayed payments and late fees.

For an accounting firm with many clients and high transaction volumes, this is a structural challenge. Staff with accounting expertise spend a large portion of their working day on manual repetition — instead of creating value for their clients.

The EHF base is large, but not fully utilized. Norway was among the first countries to require electronic invoicing (EHF/Peppol format) for all invoicing to the public sector. The infrastructure is well-developed and covers a large share of the business community. Yet many businesses receive PDF invoices by email and process them manually — even though their vendors technically support EHF. This is exactly the gap that AI integrations are now helping to close.

The cost of manual invoice processing is rarely visible as a single line item in the accounts. It hides in hours spent incorrectly, in delayed payments triggering late fees, and in accounting staff using their professional expertise on manual data entry instead of advisory work. For businesses with hundreds of monthly invoices, this quickly adds up to a structural cost — and that's exactly the cost AI invoice processing is designed to reduce.

FabricAI and Tripletex: AI for 80,000+ Norwegian SMBs

In 2024, the Norwegian AI vendor FabricAI entered into a strategic partnership with Tripletex to integrate AI-based accounts payable automation directly into the platform. The result is that over 80,000 Norwegian SMBs already using Tripletex now have access to AI reading and automatic posting without switching systems or investing in separate software (FabricAI and Tripletex, 2024).

This is a significant scale-up — and commercially successful: the joint solution generated 30 million NOK in new revenue for Tripletex in 2024 alone (FabricAI, 2024). Previously, equivalent functionality required separate integration projects, vendor agreements with ERP providers, and IT resources that most SMBs don't have. Now it's a configurable add-on in a system many already pay for.

FabricAI trains its models on Norwegian accounting practice and Norwegian charts of accounts, which provides higher posting accuracy than generic international solutions. The system learns from corrections: the more invoice data it sees, the higher the automation rate over time.

The pattern isn't unique to Tripletex. Equivalent AI layers now exist for Fiken, Visma, 24SevenOffice, and Xledger — and the adoption curve is rising as the technology matures and integrations become simpler to set up.

How AI invoice processing works in Tripletex

For an accounting firm or internal finance department using Tripletex, an AI-enabled invoice setup typically looks like this:

Reading and interpretation. EHF/Peppol invoices are received directly and in structured form. PDF invoices are interpreted by OCR combined with NLP, reading vendor data, amounts, VAT codes, and due dates. Fault tolerance is significantly better than pure OCR because the AI layer can disambiguate ambiguous text and correct formatting errors.

Automatic posting proposal. Based on vendor history, invoice category, and transaction type, AI proposes the correct account and VAT code. For established vendor relationships with clear patterns, the hit rate is very high, and a large share of invoices are processed without human intervention.

Rule-based approval workflow. Invoices above a configurable threshold amount, or with low confidence in the posting proposal, are automatically sent for manual approval. The rest are completed automatically.

Payment scheduling. Tripletex combines the due date with available liquidity and proposes payment runs. With the right setup, payment can be approved in one step and executed automatically at the right time.

Reminders and collections. For sales invoices, Tripletex supports rule-based reminders. AI layers that adapt the reminder timing and tone based on customer history are under active development and are beginning to become available in Norwegian platforms.

The Fiken pattern: same ambition, simplified approach

Fiken is designed for SMBs without an accounting background and has prioritized a simple UX. This means AI integrations are fewer and less configurable than in Tripletex — but the underlying pattern is the same.

Fiken currently supports automatic import of EHF invoices and connects to bank reconciliation via Open Banking (PSD2). Receipt apps like Expensify or Mobilexpense scan paper receipts and send structured data into Fiken, eliminating manual entry for expense reimbursements.

For accounting firms serving Fiken clients, the biggest automation gains don't necessarily come from Fiken itself, but from what you put around the system:

• Structured capture. All vendor contacts use an EHF address. Invoices arrive directly in Fiken without manual routing.
• Receipt app for expenses. Employees use a mobile app. Data flows to Fiken without manual entry.
• Approval flow. The responsible manager is notified via email or Slack and approves in the Fiken interface. Payment is scheduled automatically.
• Clear deadlines. Clear internal rules about when invoices must be processed, with AI-assisted prioritization of what's due.

This isn't "full AI" in a technical sense — but it eliminates the hunt for invoices, and that's the actual problem these businesses need to solve.

SEMINE: AI accounting built for Norwegian firms

SEMINE is a Norwegian AI accounting system built for Norwegian accounting practice, integrating with Fiken, Tripletex, and other platforms. The system automates posting, document registration, and reporting.

A documented Norwegian example: Norwegian (the airline) implemented SEMINE and achieved a 70% automation rate for its approximately 5,000 monthly incoming invoices — and cut processing time per invoice by 6 days (SEMINE / Norwegian, 2024). The improvement didn't stop at efficiency: SEMINE extracted line-level data from the invoices and gave Norwegian detailed cost tracking per aircraft, location, and airport — insight they hadn't previously had access to without manual aggregation.

The picture is confirmed by broader industry data. According to Ardent Partners' AP Metrics That Matter in 2024 report, best-in-class AP departments process invoices 81 percent faster than average, and process 2.15 times more invoices touchless — that is, without any manual intervention. For Norwegian firms and accounting departments still handling the majority of invoices manually, this is a direct measure of what is achievable. The gap between businesses that have systematically invested in AP automation and those that haven't is not marginal: it's a structural difference that affects costs, processing time, and the ability to scale without increasing headcount proportionally (Ardent Partners, 2024).

The system learns from corrections: the first time a vendor invoice is processed manually, the posting is recorded. At the next invoice from the same vendor, SEMINE automatically proposes the same posting. After a few months in operation, the automation rate is significantly higher than at the start.

For accounting firms, this is a direct capacity increase: they can serve more clients with the same team, or use freed-up time to raise the value of existing client relationships.

What AI actually delivers — and what it doesn't

It's important to have realistic expectations. AI is very good at pattern recognition and rule-based decision-making — two qualities invoice processing requires at high volume. AI in invoice processing is especially strong at:

Recognition and parsing. Reading text from invoices in various formats (EHF, PDF, images), parsing structured data about vendor, amount, VAT, and line items.

Matching and validation. Linking purchase invoices to purchase orders, identifying duplicates, flagging unusual amounts or vendor changes that deviate from history.

Posting proposals. Proposing posting based on history and rules. For high volumes of repetitive transactions with known vendors, this is highly effective and gives a high hit rate.

Anomaly detection. Flagging invoices that deviate from patterns — new bank account at a known vendor, amounts far above category average, VAT codes that don't match the vendor type.

AI is weaker at:

Judgment and complex distinctions. Whether an invoice is an operating or capital expense depends on contextual knowledge rarely documented in the system. Such assessments require an accountant.

Missing or messy data. Invoices without clear vendor references or in unusual formats give lower confidence and require manual processing.

Contractual relationships outside the system. Payment terms discussed verbally and not registered cannot be considered by AI.

Firms that succeed use AI to handle volume and free up time — and retain professional expertise for exceptions.

The pattern behind those who succeed

The firms and businesses that achieve the greatest gains from AI invoice automation share one characteristic: they treat invoice processing as a system, not as a manual task.

This means:

• Clear entry points. All invoice vendors use an EHF address or send to one dedicated address. No invoices are "on their way" to an unknown location.
• Minimal manual reading. PDF invoices are interpreted by OCR, not by a person with a keyboard. Receipts are scanned with a mobile app.
• Rule-based routing. All invoices above a given threshold go to approver A. Invoices below the amount in a given category are approved automatically. No ambiguity about who owns what.
• Visibility over status. A dashboard shows what's waiting for approval, what's delayed, and what's about to come due.
• Escalation with ownership. When AI can't match or post an invoice, it's clear who handles it — and within what timeframe.

This is as much a process project as a technology project. The most common mistake is buying AI tools without cleaning up the process first. That automates chaos — and chaos just happens faster. The best implementations start with a thorough process review: who sends invoices, through which channels, and what are the most common friction points. The automation tool is chosen after the mapping, not before.

Where do you start?

The starting point isn't choosing an AI tool. It's mapping the flow:

• Entry points. Where do invoices end up today? Email, EHF, postal mail, vendor portal?
• Processing time. How long does it take from invoice arrival to payment-ready, and what are the most common causes of delay?
• Approval. Who approves what, and what happens when they're unavailable?
• Error rate. Which vendor types produce the most posting errors or missing data?

From this mapping, you can identify the two or three steps where automation delivers the greatest impact. For most, it will be reading and posting. For some, it will be approval workflow and payment scheduling. For a few, reminders and collections will give the highest return.

With that knowledge, you can choose tools and integrate them purposefully — instead of buying a broad AI system and hoping it solves itself.

Discovery Sprint

Crunchtime offers a Discovery Sprint — a structured mapping of your invoice processing workflow and a concrete plan for which automation initiatives deliver the greatest impact in your stack.

The sprint result is a prioritized action card, not a generic report. We identify the two to three initiatives with the highest ROI and show you exactly how to implement them — whether you use Tripletex, Fiken, or a combination of tools. The sprint is suitable for firms just considering automation and for those already underway who want to ensure their next initiative is the right one.

Start a Discovery Sprint →

Sources

• FabricAI — Tripletex has chosen FabricAI as their AI partner for purchase invoice automation — web.archive.org — 2022-10-03
• SEMINE — Norwegian: AI solution SEMINE delivers immediate results in AP — semine.com — 2024-12-11
• FabricAI — AI: from an expense to a revenue generator — web.archive.org — 2025-02-28
• Ardent Partners — AP Metrics That Matter in 2024 — basware.com — 2024-05-01