6 min read
Fortnox, Visma and Vitec are all adding AI. The numbers show who gets the software, not who gets paid.
Sweden’s accounting bureaus spent years sending clients to Fortnox. They worked inside the software, taught people to use it and helped build a customer base north of 500,000. Now Fortnox is doing some of the bookkeeping itself. The bureaus made the introductions, and the platform is keeping the relationship.
This spring it launched Access for company signatories and BLINK for automated bookkeeping and reconciliation. Anna Wallman, an accounting consultant who founded Wallman & Co, told Balans, “This is a first step toward taking over our whole business.”
Fortnox admitted to “some risk of customer loss.” Fair enough, though the bureaus were describing a different kind of loss, and they said so out loud.
Picture a customer who keeps paying for Fortnox but buys less bookkeeping from their accountant. The subscription price doesn’t move. The accountant’s revenue does.
Thanks for nothing
Fortnox doesn’t need to talk anyone into moving years of financial records somewhere new. It already holds them. It connects the business to its accountant. A separate vendor selling automated invoice coding has to win the customer first, while Fortnox can put a button where the invoices land anyway.
Is that competition, or just gravity?
Wallman knows which work she fears losing. Nobody yet knows how much Fortnox will take, or whether the bureaus will move enough clients elsewhere to hurt (a rebellion only works if there’s somewhere to go).
Access isn’t use
Visma sits in a similar spot, with accounting, payroll, tax and invoicing under one roof. It can introduce AI to people who already pay it.
Reported FY2025 revenue hit €2.8 billion, up 17%. Annualised Repeatable Revenue rose 19% overall and 13% organically, and customers grew 22% to 2.4 million.
Visma also disclosed 250 live AI initiatives, Smartscan among them, which the company says cuts human involvement in document processing by 93%. That is a figure for one task, not for a finance team’s whole workload.
By H1 2026, Visma said it had added 300,000 customers and that 40% of ARR came from customers with access to embedded AI. Later material put the initiative count near 550, with 200 flagged as potentially transformational.
Look at the word “access” again. A customer can have the feature, never open it and still be counted. A gym membership works the same way, and nobody confuses it with fitness. Visma doesn’t report AI revenue separately, so these numbers show distribution, not earnings.
Management calls growth “supplemented by M&A, albeit at a lower pace,” and says the model gets more valuable as AI initiatives go live. Visma can still earn more through renewals, usage or add-on services without selling a separate AI subscription. It has plenty of ways to charge for the improvement.
Buying growth
Vitec Software Group is also putting AI into its specialist software businesses, and the sales figures look good at first glance. First glances are cheap.
In Q1 2026, total sales grew 9%. Five points came from organic growth, seven from acquisitions, and currency took three away. By Q2, total growth reached 15%, mostly thanks to acquisitions, while organic growth slowed to 4%.
Acquisitions explain most of that acceleration. Inside the existing business, management said the lift from earlier price increases was fading.
There’s an accounting wrinkle too. Two subsidiaries changed how they report transaction-based revenue, which cut reported net sales by roughly 8% against the 2025 base without touching earnings, cash flow or financial position. A smaller number doesn’t always mean less money in the bank.
Vitec still needs growth now that the price increases are wearing off. AI might help. These results don’t say whether it contributed to the slowdown or to any recovery.
| Company | What the article reports | What remains unproven |
|---|---|---|
| Fortnox | Access and BLINK extend the platform into work accounting bureaus perform. | How much work moves from bureaus to the platform, and whether bureaus move clients elsewhere. |
| Visma | Embedded AI reaches an existing customer base; 40% of ARR came from customers with access to it in H1 2026. | Access does not establish use. The figures do not isolate AI revenue. |
| Vitec | Q2 2026 sales growth accelerated, mostly through acquisitions, while organic growth slowed to 4%. | The results do not isolate AI’s contribution to growth or the slowdown. |
Meter’s running
Benchmarkit puts median net revenue retention at 108% for usage and consumption pricing, against 98% for seat-based models. The measure tracks what an existing customer group spends, expansion and losses included. It doesn’t prove the billing model caused the gap.
Ramp’s spend data, covering more than 70,000 companies, shows 14% of software companies billed for an AI feature in the past year, up from 8%. Usage-billed features rose from 24% to 36% of embedded AI spend.
The Causo retention report has gross retention falling one to three points among mid-market firms across DACH, the Nordics and the UK, along with longer procurement cycles and renewals negotiated down before signing. Buyers squeezing for better terms could explain that without AI entering the room at all.
Even the bill gets hard to read. Zero, a Helsinki CRM startup, combines per-seat charges with usage credits for its AI agents. What does a finished task cost? The seat price won’t say. Our AI agent pricing tracker looks at these differences.
What’s still missing is a Nordic buyer who can lay an old contract next to a new one and say what AI changed. The market-wide figures can’t tell that story, however confidently they’re delivered.
Quiet invoices
For Fortnox’s bureaus, the change may show up in their own invoices first. A client stays on the platform they recommended and slowly needs less of what they sell. Keeping the customer on the software no longer means keeping the work.
The bureaus built the road, and someone else is collecting the toll. The loss may never arrive as a cancellation. It may just be a quieter invoice and a call that never comes.
The billing changes are easier to inspect in the IC Lab agent pricing tracker. Our analysis of AI startup defences examines why customer systems and workflow knowledge can retain value as individual features become easier to reproduce.