Screen every project owner in days. Keep the file your NCA will ask for.
EvalLens reads every campaign application end to end — deck, business plan, financial narrative — and produces an evidence-linked screening file per project owner. Your analysts verify, your Funding Committee decides, and the record exists from day one instead of being reconstructed later.
Too slow, the deal lists elsewhere. Too fast, the file won't hold
- 25% → 1%the funnel shape serious platforms publish: a quarter of applicants get full analyst review; about one percent lists. The gap is burned analyst weeksplatform disclosures
- 2–5 wkstypical onboarding to a listing decision — while project owners shop competing platforms mid-processthe pilot measures yours
- Art. 12your authorisation file describes documented screening procedures — and NCAs check you against your own filed procedures, analyst by analystECSPR
- 5 yrsof Art. 26 record-keeping. Today, for a project owner rejected two years ago, the record is often one departed analyst's inboxECSPR
“Please provide the screening record for…” — and you export, not excavate.
The next information request stops being an archaeology project. The file exists from day one. Quotes verified against source documents — no quote, no finding. Analysts confirm every rejection; the committee makes every listing decision.
Campaign application → listing decision, in seven steps
EvalLens prepares the screening file. Your analysts verify; your Funding Committee decides.
- 01
Your screening checklist becomes the standard
Team, market, business model, financial-narrative readiness, campaign readiness — your criteria, configured once, versioned, and applied identically to every application from then on. You get: your filed procedures, actually followed.
- 02
Applications land complete
From your page or your existing flow — batch in, files out. Missing documents are flagged; project owners chase themselves. You get: intake without the hand-holding.
- 03
Every application read end to end
Deck, plan, financial narrative — every page, coverage logged. Independent AI reviewer lenses (not people — parallel reads that can't anchor on each other) assess against your criteria, evidence before conclusion. You get: application #1 and #100 on one standard.
- 04
The 75% you'd reject anyway — filtered, with the rationale documented
Weak applications surface immediately with evidence. Your analyst confirms each rejection; the full rationale is documented internally, and what you communicate to the project owner is your template and your call. You get: analyst weeks back from the non-listers.
- 05
Deep diligence starts from a prepared file
Red flags and unverified claims itemised with page references, verification checklist pre-drafted. Your KYB, AML, legal review and Art. 5 checks stay exactly as they are — they just start at week three. You get: shorter weeks, same depth.
- 06
The Funding Committee compares like with like
Every candidate in the same format: evidence-linked brief, open questions, comparable grounds. Decisions, overrides and reasons logged in committee mode — exportable for the minutes. You get: faster committees, documented decisions.
- 07
KIIS completeness, checked systematically
A logged, section-by-section completeness check against Annex I that slots into your Art. 23(11) procedures. Correctness and clarity remain your review — now starting from a flagged draft instead of a blank page. You get: one limb of Art. 23(11) systematised, honestly scoped.
The paragraph your Compliance Officer reads first.
Built to ESMA’s dialect: human oversight, testing, firm accountability.
ESMA’s guidance on AI in investment services doesn’t ban AI — it demands governance: board-level understanding, human oversight, and the firm taking full responsibility for systems it deploys. That’s this architecture, literally. And the consistency claim — “the same standard for every application” — is falsifiable, not marketed: versioned criteria, versioned models, logged reads, and quote-verification against source text.
What the parallel pilot measures on your inbound.
Thirty recent applications, run side by side with your analysts — under a signed DPA, EU processing, success metrics agreed before we start.
Analyst-hours / 100 apps
Before vs after — the number your CFO turns into the business case.
First-decision SLA
Days from application to a documented first screening decision.
Claims caught
Unverified or inflated claims flagged vs the analyst baseline on the same files.
File completeness
Can every decision be answered from the standing record? Counted per application.
Fixed-fee pilot, then continuous screening priced to your monthly inbound, in EUR, invoiced. Applications are read in the language they arrive in, Nordic and Southern-European included, with scores comparable across languages. See pricing or book a call.
What compliance will ask
- Can a regulated CSP use AI in screening at all?
- Yes — under governance. ESMA's guidance on AI in investment services expects human oversight, testing, and the firm retaining full responsibility. In this design your analysts confirm every rejection, your committee makes every listing decision, and the whole process is logged. It's easier to defend than undocumented analyst-by-analyst variance.
- Does this satisfy our Art. 5 / Art. 23(11) duties?
- No tool can — and we won't claim to. Art. 5 checks and your verification stack stay yours; on the KIIS we systematise the completeness limb of Art. 23(11) against Annex I, while correctness and clarity remain your review. What you gain is documented consistency and a standing file.
- What happens when the AI misreads a document?
- Quotes are verified against the source before a finding stands — no quote, no finding. Thin evidence moves conclusions down, never up, and everything material routes to the analyst's verification checklist rather than into a silent score.
- Which model providers see project-owner documents?
- Disclosed in the sub-processor list in your security pack, with a contractual no-training commitment and EU processing options. On the Enterprise track, screening runs on your approved models — including inside your environment.
- How does it integrate with our application flow?
- Batch-based to start — export from your current intake, files and evaluations back — with API integration scoped on the Platform tier. No project-owner-facing change is required on day one.
- Who else runs this?
- We're onboarding a founding cohort of platforms, which is why the entry point is a measured parallel pilot on your own inbound — with metrics your Compliance Officer signs off before anything starts.
Thirty applications. Side by side.
Start small: through August 31 a retro-run on up to 10 past applications is free. If the file holds up, the next step is the fixed-fee pilot — one month of inbound, around thirty applications, run in parallel with your own screening under a signed DPA.