How Ephelants turned 20,000+ LinkedIn contacts into 3 ICP-ready databases with Claude

Client

Ephelants is a creative production studio building microdramas, branded content, AI-integrated video, and PR campaigns. The team also incubates Village - a separate venture currently raising seed-stage capital. Two priorities. One network. ~22,000 LinkedIn connections built across decades in advertising, film, and tech.

A network that big should have been an asset. Instead it sat invisible.

Challange

Most founders build a network for years and then can't actually use it.

Ephelants had three live use cases pulling on the same network at the same time:

1. Selling production services to brands and agencies (Ephelants client acquisition)

2. Raising capital from super-angels and seed VCs (Village fundraise)

3. Surfacing problem-solvers when partners or peers ask "who do you know who can...?"

Manually scrolling through 22,000 connections to find a CMO at the right agency, or a seed investor with media-sector fit, or a documentary cinematographer for a referral - that's a few hours of LinkedIn searching every time. Multiply by 5 lookups a week. Compound over a year. The math is brutal.

And the right contact often sat four connections deep in a LinkedIn screen, then got buried again before anyone could act on it.

He had the network. He just didn't have the infrastructure to use it.

Our hypothesis

Build the database once. Query it forever.

Founders don't fail at prospecting because they lack contacts. They fail because they can't surface the right one when it matters. Solve retrieval and you compress weeks of work into seconds - and find people you'd genuinely forgotten you knew.

Our solution

Three steps. Two weeks. One database that lives forever.

1. Extract & qualify

We pulled all ~22,000 LinkedIn connections, removed duplicates, dropped expired profiles, and applied qualification logic. What stayed had active roles, complete profiles, and sector relevance to one of the three priorities. Final count - 19,441 enriched contacts.

2. Enrich & label

We applied a custom taxonomy built around the actual go-to-market reality:

- Buying power - decision level (C-Suite, VP, Director) and budget authority

- Relationship strength - first-degree, mutual connections, recency of interaction

- Sector fit - mapped to Ephelants ICPs (brands, agencies, studios) and Village ICPs (super-angels, seed VCs, sector operators)

- Connector ability - whether this person can introduce you to the next layer of network

Many contacts ended up in two segments. A CMO at a media holding company can be both a potential Ephelants client and a Village investor signal at the same time.

3. Set up Claude as the query layer

We packaged the database with a system prompt that teaches Claude how to read the taxonomy. Open claude.ai, paste the system prompt, attach the database file, ask in plain English. No SQL. No filters. No spreadsheet wrangling.

How it works in practice

A few prompts away from the right person.

Claude recommends. The founder decides. No automated outreach, no auto-sending - the human stays in the loop on every contact decision. The system surfaces options. The founder picks who to actually message.

What this actually unlocks

A founder's network is the most underutilized asset in early-stage business. Not because it's small. Because it's invisible.

You can't use what you can't surface. Once the database exists, every new conversation, every intro request, every fundraise round runs faster. The work is done once. The compounding starts immediately - and it doesn't stop.