The PR Coffee Party Pipeline (How a Startup Got 4 Features Without an Agency)

PR agencies typically charge $5,000 to $15,000 per month. For an early-stage startup, that’s a real cost — not just in dollars but in the opportunity cost of spending that budget on something that might generate coverage or might not.

Lucas Siegel, the founder of Yuna (an AI mental health platform with $30M valuation, 50,000+ users, and partnerships in progress with Amazon and T-Mobile), had a different approach. He built an automated PR pipeline over a weekend and got featured in four Austin news outlets. No agency, no retainer.

He calls it the “coffee party.”

Why Most PR Pitches Fail

Before getting into how the pipeline works, it’s worth understanding why most PR pitches don’t land.

The typical PR approach is spray-and-pray. You write a press release, compile a media list from a database or past contacts, and blast it out to as many journalists as possible. The pitch is generic. The press release is written in the third person and reads like a legal document.

Journalists receive dozens of these a day. Most get deleted before they’re even opened.

The ones that get read — and the ones that turn into features — share something in common: they feel personal. They show that the sender actually read the journalist’s work. They connect the story to something the journalist has already been covering.

The principle is what Lucas calls the coffee party model. Instead of treating a journalist like a distribution point for your press release, you treat them like someone you’d want to grab coffee with. You’d never walk up to someone at coffee and pitch them without knowing anything about them. But that’s what most PR pitches do.

The pipeline automates the research and personalization that makes the coffee party approach work at scale.

Step-by-Step: How the Pipeline Works

Step 1 — Journalist Research

The pipeline starts with ChatGPT deep research to identify journalists who are actually relevant to your story.

For Yuna — an AI mental health platform targeting HR professionals and enterprise buyers — the relevant journalists are covering mental health, workplace wellness, Austin business, HR tech, and AI applications in healthcare. That’s a specific set of people, not everyone on a general tech press list.

ChatGPT identifies who’s been writing about these intersections, which publications they work for, and what their recent coverage has focused on.

Step 2 — Data Enrichment

Once you have the journalist list, a Lindy agent takes over. It scrapes email addresses and professional contact info, Twitter/X handles, and their articles from the last 3-6 months — specifically what they covered, what angles they explored, who they quoted.

This takes a few minutes automated. Manually, this is the step that PR agencies bill hours for.

Step 3 — Personalized Outreach

Here’s where the approach diverges from standard PR.

Each email opens with a direct reference to something the journalist actually wrote. Not a compliment. Not “I’ve been following your work” (which signals you haven’t). A specific reference: “I saw your piece on workplace burnout at tech companies last month…”

Then it connects the company’s story to that angle. If the journalist covers workplace mental health, the pitch is framed as a workplace mental health story. If they cover Austin startup growth, it’s an Austin startup growth story.

The insight here: editing is faster than authoring. The journalist doesn’t need a generic press release — they need something they can quickly understand and connect to their beat. An email that does that work for them is dramatically easier to respond to than one that makes them figure out why they should care.

Step 4 — Conditional Follow-ups

A branching follow-up sequence handles the scenarios that kill most outreach campaigns:

• No open after 2 days: Send a different subject line. A second subject line gives you another shot.
• Opened but no click or reply: Send a short “just wanted to make sure this didn’t get buried” follow-up. This captures people who were interested but got distracted.
• No response after follow-up: Let it go. Journalists who don’t respond after two attempts are probably not interested right now.

Result: Four features in Austin news outlets.

The Economics

At a traditional PR agency, the hourly rate for the research and personalization step alone — identifying relevant journalists, reading their work, crafting personalized pitches — would easily run $200-400/hour. For a list of 50 journalists, you’re looking at many hours of work.

The automated version of those same steps runs on AI API calls that cost cents.

The quality of personalization is arguably higher in the automated version, because the research step is instant and the agent can pull more context per journalist than a human researcher would typically invest time in.

The part that remains genuinely human — and should — is the story itself. The angle. What makes this company interesting enough to write about. What connects its work to what journalists are already covering.

That’s the part the pipeline can’t do for you. But once you have a story worth telling, the pipeline does the distribution work.

Building Your Own Version

If you want to replicate this for your company or a client, here’s what you need:

1. A clear story angle — what’s actually interesting about this company, product, or announcement?
2. A journalist research step — ChatGPT deep research works well here. Give it your industry, the angles you’re pitching, and ask it to identify journalists who cover those intersections.
3. A data enrichment agent — Lindy, Clay, or a similar tool to find email addresses and pull recent articles.
4. An email template with personalization fields — the core template stays consistent, but the opening line adapts to each journalist.
5. Conditional follow-up sequences — one follow-up is appropriate, two is pushing it, three is probably spam.

The Broader Principle

There are tasks that require genuine human judgment (writing the story, deciding whether the angle is compelling) and tasks that are research-intensive but don’t require judgment (finding journalists, reading recent articles, pulling contact info). AI is genuinely good at the second category.

The key insight: editing is faster than authoring. When a human reviews and approves AI-drafted personalized emails instead of writing from scratch, they can process 10x more outreach in the same time without sacrificing quality.

For PR specifically: the research and personalization step is what separates effective outreach from spam. AI makes that step nearly free. What you do with the time you save is up to you.