From Typeform to Decision: Automating Application Intake, CRM Sync and AI Scoring for PF Network
The Problem
PF Network is an invite-only community for European tech founders, investors, researchers and policymakers. Every applicant completes a dynamic Typeform: founders, investors, researchers, policymakers and students each see different questions. Responses were landing in spreadsheets, so the team had no structured view of who was applying. There were three specific problems: More than 200 existing applications sat in a CSV export, much of it corrupted by character-encoding errors. Answers weren’t reaching the Client CRM, so every new application meant manual copying. Reviewing applicants meant reading each one individually, with no consistent way to rank them against the Network’s membership criteria.
The Solution
I built an end-to-end automation in n8n, connecting Typeform, the Client CRM, their newsletter platform, Google Sheets and Claude. CRM structure: a setup workflow created around 40 custom fields in the Client CRM through the API, so the data model was consistent from day one. One-off import: an import workflow cleaned the historical data, repaired the encoding damage, and collapsed Typeform’s one-column-per-checkbox exports into proper multi-select fields. Live sync: a Typeform trigger sends every new submission to the Client CRM in the same way. People are matched by email and companies by domain, so a repeat applicant updates their record instead of creating a duplicate. Dropdown options are created automatically when a new answer appears. Newsletter: applicants who consented to marketing are subscribed to the PF newsletter in their newsletter platform. The workflow checks their subscription status in the Client CRM first, so anyone who previously unsubscribed is never re-subscribed. AI scoring: Claude Haiku scores each application against a tiered rubric, for example exited founders, then funded founders, then investors, with automatic rejection of out-of-scope applicants. The score and a short written rationale are written back to the Client CRM record. To agree the rubric with the client, I built a small interactive tool that walked them through weighting each question and describing strong and weak answers, then generated the scoring prompt. Reporting: every application and its score is also logged to a shared Google Sheet for review outside the CRM.
The Outcome & Value Delivered
200+ historical applicants imported into the Client CRM as clean, structured records, with new submissions arriving automatically from then on. Every applicant now arrives with a score and a plain-English rationale, so the team can start from a ranked shortlist. Newsletter subscription and unsubscribe preferences are respected automatically. The client has the rubric as an editable tool, so scoring criteria can change as the Network’s membership priorities evolve.