The EU AI Act Goes Live in Two Weeks. Most Finance Teams Are Not Ready. Neither Was I.
August 2 is the date. That is when the EU AI Act’s transparency obligations and conformity requirements for high-risk systems become fully applicable. Not proposed. Not phased. Applicable.
The headline number from AlixPartners, published this month: only 15 to 20% of European companies are prepared for this transition. In financial services, the number is not meaningfully better. The regulation affects every firm using AI in credit scoring, investment decisions, client risk classification, or document processing that feeds into those decisions. That covers most of what a PE fund, credit team, or M&A boutique does with AI today.
The compliance gap is not about awareness. Every partner at every fund has heard the AI Act mentioned in the past six months. The gap is operational. What tools are processing client documents. Where those documents go. Whether there is an audit trail. Whether the model can be explained to a regulator if asked. Most firms using AI in a deal workflow cannot answer all four questions cleanly, because the AI they are using was not built with those questions in mind.
The pattern is consistent. A firm adopts a general-purpose AI tool because it is fast, accessible, and good enough. It sits outside the firm’s governed stack. No audit logging, no data residency documentation, no explainability layer. Nobody thought it was permanent infrastructure. It became permanent the week it started saving the team two hours per deal.
Then the compliance question arrives and the answer is: we are not sure what went where.
The European landscape shows a stark pattern: approximately 68% of tech startups have adopted AI, but only 53% of large traditional enterprises have, and just 3% of those large companies have integrated AI into their core operations in a governed way. For mid-market finance firms sitting between those two categories, the exposure is the sharpest. Enough AI usage to create regulatory surface area. Not enough infrastructure to cover it.
The EU AI Act does not require firms to stop using AI. It requires them to know what they are using, document why it is appropriate for the use case, and demonstrate that the system behaves consistently and explainably. For a high-risk use case and document analysis feeding investment decisions sits in that territory the bar is specific.
Most of the AI tools currently running in European finance deal workflows do not meet that bar out of the box.
I have been building Lens for the past several months to close exactly this gap for mid-market finance teams. Document intelligence built for European regulatory requirements from the start. GDPR-native, EU data residency, full audit logging, citations on every output so that any answer can be traced to its exact source in the original document.
The timing of shipping this two weeks before August 2 was not planned around a compliance calendar. But it is not accidental either. The reason EU data residency and explainable outputs are core architecture rather than features on a roadmap is that I was building for an environment where regulators would eventually ask these questions. That environment arrived faster than most expected.
Lens is live today.
The product: a document intelligence platform for PE funds, M&A boutiques, credit teams, and legal practices in the European mid-market. The core interface is Matrix: documents as rows, questions as columns, AI agents filling each cell in parallel, every answer citing the exact page and sentence in the source document. Deal Library for institutional memory that survives analyst turnover. Thesis Builder for investment theses connected to live market data. Origination Signals EU for sourcing deals before they go to formal process.
Pricing starts at 99 euros per month. The team tier, which is the one most boutiques will use, is 299 euros per month for ten users and unlimited documents. No custom implementation. No six-month sales cycle. Eight minutes to get started.
For finance professionals: If your team uses AI in any part of your deal workflow and you have not done a clean inventory of what tools are touching what documents and where that data goes, August 2 is a deadline worth treating seriously. Happy to share what a basic compliance inventory for a mid-market fund looks like in practice, no product pitch involved.
For builders: I built this solo over six months, mostly with Claude Code, while finishing a finance degree and starting an internship at a bank in Luxembourg. No co-founder, no marketing budget, no paying customers yet. The goal for the next eight weeks is five. If you have built something in B2B SaaS and have opinions on what that path looks like, I want to hear them.
