Why kvit - a founder letter

My previous company was a German receivables business. Built the classical way: BfJ-registered, close to a hundred people at peak, a book of thousands of overdue B2B and B2C receivables per month, a team of caseworkers running the escalation stages. We grew over the years to eight-figure annual revenue - solid by the industry's standards, slower than our own ambition.

What those years gave me was the inside view you can't get from the outside. Thousands of dunning sequences. Hundreds of phone negotiations. Court Mahnverfahren proceedings. Conversations with lawyers about enforcement titles. And across those years a single observation kept hardening: eighty percent of what the caseworkers did was process repetition - routine work that a human was doing only because no software was good enough to take it over.

I spent a long time trying to solve that from the inside - buying software, building software, tightening processes. It worked partially, because the business model itself depended on human handling: the success fee was tied to human work, not to software output. At some point it became clear that this isn't an optimisation problem inside the old model. It's a new business.

Three problems that the classical model cannot solve from within itself:

Incentive misalignment. Classical success fees run 15 to 35 percent of the recovered amount. That incentivises maximum pressure, not relationship preservation. Anyone who knows the business knows: most overdue B2B receivables are not bad faith, they are cash-flow gaps or stuck internal approvals. Maximum pressure is the wrong lever in exactly those cases - it damages the creditor's customer relationship without materially raising recovery probability.

Scaling problem. Every new case in the classical model needs a human caseworker. Margin scales linearly with payroll. For German SMEs with twenty open items a month, the maths is bad - the fixed handling fee eats the small receivables, while the large ones generate the firm's profit.

Language and channel. A classical collection firm has German-speaking caseworkers and sends letters and makes phone calls. In a B2B market where debtors are increasingly international and prefer different channels (email, WhatsApp, SMS, voice), monolingual written-only handling is a systematic bottleneck - and not one solvable with more staff.

These structural problems have existed for decades. What makes them solvable now is three movements, all of which crossed maturity inside the last 24 months:

Voice AI is productive. Models from ElevenLabs and Vapi reach a voice quality, latency, and language range in 2026 that work in real telephone negotiations. An overdue French debtor called by a fluent French voice reacts fundamentally differently than to an English or German call. Two years ago that was science fiction. Today it is a product.

ViDA and the e-invoicing mandate make incoming data structurally available from 2025. XRechnung and ZUGFeRD become the normal form; the manual PDF processing that today still inserts seven to fourteen days into every handling flow disappears. Whoever is built on structured invoice data benefits immediately.

The EU AI Act brings clarity, not weight. What looks like regulatory pressure is, for a cleanly built software-native solution, actually market cleanup. Whoever scores natural persons (Annex III §5(b)) gets hard obligations, justly. Whoever scores invoices - a contractual document, not a person - falls outside the high-risk category. That isn't a gap in the law, it's a deliberate distinction. We placed kvit on that distinction from the first commit.

kvit is the operational layer that sits between your billing system and your overdue customers. Concretely:

Integration in 60 seconds with Stripe, Lexoffice, or Sevdesk via OAuth. As soon as an invoice goes overdue, kvit takes over - drafts the Mahnung in the debtor's language, escalates across the right channels, calculates default interest and the flat fee, documents end to end, and hands off when needed to a BfJ-registered partner for the court Mahnverfahren.

The score we attach to every invoice tells you where the lever is biggest: which invoice will likely be paid on its own, which needs a gentle reminder, which is already heading for escalation. That score belongs to the invoice, not the person - deliberately, EU AI Act-compliantly.

Pricing: free tier up to five open receivables per month; Pro at EUR 49 plus 4 percent on actually recovered amounts; Legal tier with the partner for court proceedings. No hidden RVG fees, no minimum term, cancellable monthly.

Just as important as what we are building is what we deliberately are not doing:

Not a new collection agency. We build the software the creditor operates themselves. That keeps us out of RDG legally, out of the success-fee spiral economically, and out of the escalation-by-default logic culturally.

Not a US-first product with DACH translation. kvit is built from the first commit for German, then Austrian and Swiss SMEs, with the DACH specifics baked in from day one: the EUR 40 flat fee under § 288 (5) BGB, ZUGFeRD and XRechnung formats, correct Mahnstufen logic, EU data residency. US markets we will serve later, not first.

No AI-for-everything promise. We use LLMs at exactly three places: language of the Mahnung (tone matched to context and debtor profile), voice scripts for outgoing calls, score on the invoice. Everything else is deterministic software. If a step can be expressed predictably, we don't want to make it a black-box LLM call.

The name comes from Norwegian: kvitt means paid, settled, free of it. There is no clean equivalent in German - erledigt comes closest but is operational, not emotional.

The word choice is a statement. The old industry words - Mahnung, Forderung, Inkasso, debt collection - describe the process from the perspective of pressure. They mirror what is being done, not what is being achieved. Kvit describes the outcome: the invoice is paid, the customer has their status back, the relationship continues.

That's not branding cosmetics. It mirrors a different posture toward overdue receivables - one the old business taught me, but could never embody, because its revenue model was built on pressure rather than resolution.

This note isn't addressed to investors, not to competitors, not to the trade press. It's addressed to the finance leads, the bookkeepers, the SaaS operators, and the trades who at the end of every month have one open item too many in their books and don't want to engage a collection firm to deal with it.

If that resonates: kvit is in closed beta. Free tier is connected in under a minute, no contract. Direct contact goes to hello@kvit.ai - we read every message personally, at least while we are still small enough that we can.

If it doesn't resonate: the next posts get technical. Default interest, Mahnung templates, the SCHUFA question. If you landed here from a search query, the technical answers are in the cross-references below.