Case studies

Sector

Financial Services

Engagement

Inbox triage & matter routing

Stack

Gmail · OpenAI · N8N

Live since

May 2026

Investment team members were spending the first 30–60 minutes of every day cutting through noisy inboxes, where deals and actionable threads were easy to miss. We built a quiet classifier that reads inbound mail, applies the right tag inside the inbox, and surfaces what matters first. Less time on organizational triage, and nothing important gets lost in the noise.

• 2+ hrs reclaimed at every member · per week
• 20% drop in misrouted matters · 30-day window

We stopped fighting our own inbox. The system handles the classifying and we focus on the emails that matter.

Sector

Financial Services

Engagement

AI deal-flow triage

Stack

Claude Sonnet · n8n · Postgres · Tavily

Live since

May 2026

Pre-seed and seed funds drown in inbound pitch decks; partners read maybe one in five and still miss the strongest ones. We built an end-to-end agent that takes the inbox off their plate. Founders submit through a branded public form, the system extracts everything findable from the PDF, web-researches each founder independently, scores the submission against the fund's own thesis, and drafts a partner-ready outreach email, all before the partner opens their morning coffee.

• ~90 sec end-to-end per submission · upload to scored verdict
• €0.20 all-in cost per submission · LLM + web search
• 6 filters hard thesis check before any LLM cost is spent
• 100 % human-in-the-loop · outreach emails are drafts, never auto-sent

The fund's thesis is the central config: hard filters (founder equity, round size, pre-money cap, geography, business model, constitution year) plus a free-text qualitative thesis fed to Claude for nuanced scoring. Swap the config and the same pipeline serves a different fund.

Demo is gated · ask us for credentials.

Sector

Professional Services (SME)

Engagement

Invoice & receipt extraction → bookkeeping

Stack

Gmail · Claude (vision) · Holded · n8n

Live since

April 2026

The office manager spent six hours every week on the same loop: open the supplier email, save the PDF, type vendor, amount, IBAN, VAT, and cost-centre into Holded, file the email. We hooked Gmail into a vision-capable agent that does all of that automatically, and only surfaces low-confidence items for a quick daily review. Same accounting software, same supplier emails, six hours back.

• 6 hrs / wk reclaimed for the office manager · ≈ 25 days a year
• < 2 % items needing manual review · after first 30 days
• €0.04 all-in cost per invoice · LLM + vision
• 1 week kickoff to live · no migration, no new tools

Same template applies to expense receipts, supplier statements, purchase orders, and shipping confirmations: anywhere an SME's ops team is the human OCR for a recurring document type.

Sector

Venture Capital

Engagement

LP discovery & fit-scoring

Stack

n8n · LLM · Supabase · Clay/Apollo · Tavily

Live since

2026

Raising a fund means finding the right limited partners, and most GPs still do it from memory and scattered spreadsheets. We built an agent that discovers net-new prospective LPs from public sources, enriches each one, and scores fit against the fund's own thesis, handing partners a ranked, reviewable shortlist instead of a blank page.

• 100 % human-in-the-loop · stops at a reviewable list, never sends outreach
• fit × warmth priority score · thesis fit weighted by warm-path strength, computed in Postgres
• 1 config swap the fund thesis and the same pipeline serves a different fund