FUNDING DESK · Sweden · AI compliance and third-party risk management (RegTech / GRC), aimed at regulated enterprises and smaller companies that supply large customers.
Proofbacked
Proofbacked is an AI compliance platform where the AI prepares supplier assessments, framework mappings and questionnaire drafts, while a named person makes and evidences every decision.
Company website ↗First-party material retrieved
Last retrieval:2026-11-21
Evidence gaps:Pricing · Customer cases · Technical documentation
Much of the manual work in a compliance team: reading contracts and certificates, reconciling supplier statements, mapping frameworks such as NIS2, DORA and ISO 27001, drafting customer questionnaire answers and chasing missing evidence, plus part of the first-pass triage done by external vCISO or advisory firms.
Its pitch shifts from “AI does compliance” to “AI prepares, humans decide, decisions are provable,” which fits the auditability constraint of compliance work; whether it truly replaces consultant or headcount cost remains unverified by the available evidence.
The following is editorial analysis based on public material. Inferences and open questions are labeled in the text. Funding is not evidence of revenue or product-market fit.
Product evidence checked 2026-11-07
01
What the product does
The site splits the product into Third-Party (supplier review, tiering, contract and certificate comparison), Programme (NIS2, DORA, ISO 27001, GDPR, CRA and AI Act treated as one set of controls, owners and evidence) and Customer responses (import a customer questionnaire and draft answers from reviewed facts), all sharing one context of policies, facts, requirements and promises. The product explicitly removes approval, tiering, sealing and external sending from the AI and reserves them for a named person, turning each decision into a shareable, tamper-evident receipt. The contract review, framework matching and questionnaire drafting shown on the site are labelled illustrative or fictional, so real depth, integration coverage and accuracy are not independently verified.
02
Users, buyers and demand
Users span sales, procurement, security, compliance and the board: sales needs fast questionnaire answers, procurement chases supplier evidence, security rechecks controls, compliance leads map new frameworks, and management must explain who decided what on which evidence. The buyer is most likely the compliance or security lead or CISO, and the pain recurs frequently — every new customer questionnaire, new regulation or annual review forces the work again. The site does not disclose customer size, industry mix or paying role, so this is inferred from the described roles.
03
The actual workflow
Inputs are supplier contracts, certificates, supplier statements, internal policies, regulatory texts and customer questionnaire files; the AI reads and compares them, flags which conclusions have source support and where sources conflict (for example a supplier statement claiming EEA data residency against an audit report saying otherwise), and stops where judgment is required. A named owner reviews sources, confirms matches, records reasoning and decides, and the system generates a receipt with decider, mandate, evidence links and timestamp. Missing evidence returns to the owner for follow-up, and a new framework can be dropped in to show how much existing work already covers and what is new. The key design choice is keeping sources verbatim rather than treating a summary as the whole truth.
04
Pricing and unit economics
The site offers only “Book a demo” and discloses no pricing, billing unit (per supplier, per seat or per framework) or free tier, so unit economics cannot be inferred. It is reasonable but unverified to infer a B2B SaaS subscription anchored on replacing compliance headcount; whether that anchor holds depends on actual deployment and human review time, and the evidence does not support gross margin or acquisition cost estimates.
05
Adoption evidence and gaps
Confirmed adoption signals are limited to the company saying it is out of stealth, three named executives (CEO, CCO, CISO), Swedish corporate details, and claimed integrations with Fortnox, Visma, Microsoft 365, Entra ID, GitHub and Jira. All customer cases, performance figures and compliance conclusions on the site are examples or company-reported, with no named customers, retention, revenue or audit opinion; the funding record gives no customer count either. Adoption evidence is therefore thin.
06
Competition and defensibility
Against general-purpose models, the differentiation is not model capability but product constraint: “the AI may not approve, seal, sign or send” is enforced in the product rather than left to a prompt, and every decision traces to sources and mandate, which fits audit and regulatory expectations of explainability. Against legacy workflows (spreadsheets, email, consultants), it makes multi-framework mapping and evidence reuse a shared foundation. The moat condition is whether auditors actually accept the receipts, how deeply framework libraries and integrations are maintained, and whether switching costs rise as evidence accumulates; all of this is currently a design claim, not externally validated.
07
How to read this round
This is an Other-stage round with amount and currency rendered by the deterministic layer; the investor list includes Saminvest, Skåne Ventures and two individuals, but a list is not a lead and the round should not be guessed as Seed or Series A. The financing indicates early capital to continue product and go-to-market work, but it is not evidence of demand validation or product success, and use of funds is undisclosed. The next milestone should be publicly verifiable named customer cases, a pricing page and documentation rather than more feature claims.
08
Where it could fail
First, if AI preparation accuracy is insufficient or source comparison frequently errs, reviewer overhead cancels the “no new hires” value proposition and renewals fail. Second, compliance procurement is slow and trust-heavy; if auditors or regulators do not accept the receipt mechanism, the product degrades into another document management tool. Third, frameworks such as DORA, NIS2 and the AI Act keep changing, and maintaining libraries and integrations may strain a small team. Fourth, general model vendors or incumbent GRC players could build similar human-in-the-loop capabilities and compress the differentiation.
09
What you can take from it
The most transferable lesson for AI product builders is the boundary design: rather than letting the AI make final decisions, it separates preparation from judgment and turns human decisions into verifiable product assets. That makes it easier to sell into high-trust compliance settings than a fully automated narrative, and each use accumulates switching cost.
10
What to watch next
Verifiable next indicators: whether a pricing page and product documentation appear, whether named customer cases or auditor acceptance of receipts emerge, whether integrations move from claimed to documented read-only connections, and what the next round’s stage and use of funds disclose.