Case study20 October 20265 min read
Case study: how Layer Finance got a working AI extraction pipeline in three weeks
Six months of in-house work had not reached the quality Layron needed. Three weeks later a prototype outperformed it, and Layer Finance measured 80% time saved in deal onboarding.
By Michael PresecanAIgenica
Client: Layer Finance (product: Layron), institutional real estate, offices in Baar, Switzerland and Vienna, Austria. Result: working prototype in 3 weeks, outperforming six months of in-house work. Layer Finance measured 80% time saved in deal onboarding.
The situation
Layer Finance sources off-market real estate deals, arranges debt and equity and structures real estate securities. Their product, Layron, turns the documents of a deal into investor-ready briefs: fact sheets, deal scores and investment memoranda.
Everything in that chain depends on one step: extracting the right facts from the deal documents. An error there flows into every output after it, including the deal score an investor looks at first.
The extraction had been in development in-house for six months and had not reached the quality the product needed. The rest of the roadmap was waiting for it.
What we did differently
We did not start from an empty repository. We started from a proven foundation, so the first day already had sign-in with roles, separate organizations for each client firm, a typed data model, an admin panel to review and audit every AI output, and a dashboard with billing.
That meant the three weeks went where the value was:
- Understanding the documents the way the team does. Which facts matter for a deal, where they hide in a data room, and how an analyst checks them.
- Building the extraction pipeline around those rules, with every result traceable to its source document.
- Making review easy. The team can see, correct and approve what the AI extracted, so quality improves instead of silently drifting.
The result
After three weeks the prototype outperformed the existing extraction pipeline and covered more of the product than the previous six months of work. It was built on a conventional, documented stack, so the client's own team could take it over, and Layer Finance moved its entire solution onto it.
We measured 80% time saved during deal onboarding. — Kurt Praszl, Head of Investment, Layer Finance
Since then Layron has grown into a deal cockpit with data room ingestion, agentic due diligence, deal scoring, and fact sheets and investment memoranda generated from each firm's own templates.
What other businesses can take from this
- Start from how experts already work. The extraction got good because it followed the checks analysts already do.
- Make AI output reviewable. An admin view where people approve and correct results is what makes automation trustworthy.
- Don't rebuild the basics. Authentication, roles and billing are solved problems; spend the time on the part that is unique to you.
The 80% figure was measured and published by Layer Finance.
