Automate PR Pitch personalization and Media List Segmentation

Most PR teams spend somewhere between 40 and 120 hours preparing a single media campaign. Not because the work is conceptually hard, but because the workflow is a grinding loop of research, cross-referencing, writing, rewriting, and praying that the journalist you just spent 45 minutes researching hasn't switched beats since last Tuesday.

The response rate on a generic blast pitch hovers around 3–5%. A genuinely personalized pitch—one that references a journalist's recent work, matches their angle preferences, and connects your story to something they actually care about—lands closer to 30%. Everyone knows this. The problem has never been awareness. It's been time.

This is a guide to building an AI agent on OpenClaw that handles the research-heavy, repetitive portions of PR pitch personalization and media list segmentation—without producing the kind of detectable AI slop that gets you publicly shamed on journalist Twitter. We'll walk through exactly what the manual workflow looks like today, what an agent can realistically take over, what still needs a human, and how to wire the whole thing together.

The Manual Workflow: What You're Actually Doing for 80 Hours

Let's be honest about where the time goes. A typical PR campaign targeting 50–100 journalists involves these steps:

Step 1: Media List Building (2–8 hours)

You open Muck Rack or Cision. You search by beat, outlet type, and topic keywords. You cross-reference with Twitter/X and LinkedIn because no single database is complete. You export a list of 80–150 names, knowing 20–40% of the contact data is stale because journalists change jobs every two to three years on average.

Step 2: Deep Research Per Contact (10–45 minutes each)

For every journalist worth pitching, you read three to ten recent articles. You're looking for their angle tendencies, the topics they return to, the framing they prefer, whether they're skeptical or enthusiastic about your space. You check their social accounts for recent commentary, job changes, or editorial calendar hints. You note if they've covered competitors.

For 75 journalists, this alone is 12–55 hours.

Step 3: Personalized Angle Development (10–20 minutes each)

You connect your client's news to the journalist's specific interests. "You wrote about supply chain transparency gaps in March—our blockchain pilot addresses exactly the vulnerability you identified." This is the part that actually moves response rates. It's also the part most teams skip when they run out of time and start copy-pasting.

Step 4: Writing and Customization (15–30 minutes each)

Unique subject lines. Custom opening paragraphs. Adapted data points, story angles, and suggested headlines. Multiply by 75 contacts and you're looking at another 18–37 hours.

Step 5: Review, Send, Track

Senior review for tone and accuracy. Manual entry into your sending platform. Follow-up scheduling. This is administrative, but it's real time.

Total: 40–120 hours per campaign. One full-time person, fully dedicated, for one to three weeks. For a single campaign.

Why This Is Painful Beyond Just the Hours

The time cost is obvious. The less obvious costs are what actually kill you:

Data decay ruins your personalization. You spend 30 minutes researching a journalist only to discover—after you've sent the pitch—that they left that outlet two months ago. Muck Rack's own data shows databases carry 20–40% inaccuracy rates. Your carefully personalized pitch lands in a dead inbox or, worse, reaches someone who now covers an entirely different beat.

Surface-level personalization backfires. Referencing an article from eight months ago that a journalist has moved on from doesn't read as "personalized." It reads as "I searched your name and grabbed the first result." Journalists in 2026 and 2026 have become extremely good at spotting this. Muck Rack's 2026 State of Journalism Report found that 82% of journalists say they're more likely to respond to pitches referencing their recent work—but only 37% of pitches they receive are meaningfully personalized. The gap is enormous.

Volume and quality fight each other. You either send 200 generic pitches and get a 4% response rate (8 responses, most lukewarm), or you spend 60 hours personalizing 50 pitches and get a 30% response rate (15 responses, most engaged). The math favors personalization every time. But most teams can't afford the time, so they default to volume and accept the worse outcome.

Creative fatigue is real. Reading your fortieth cybersecurity article in two days to find a unique hook for journalist number forty-one is genuinely draining. Quality drops precipitously after hour six of research in a single sitting.

The cost structure doesn't scale. A junior PR coordinator costs $50K–$70K annually. A senior strategist costs $90K–$140K. Using senior time for research that could be automated is an expensive misallocation. Using junior time without senior oversight produces bad pitches. Neither option scales.

What an AI Agent on OpenClaw Can Actually Handle

Let's be specific about what's automatable and what isn't. The industry is in what some people call a "centaur" phase—the best results come from humans and AI working together, not from full automation. The key is knowing exactly where to draw the line.

High-confidence automation (OpenClaw handles this well):

Research aggregation: Ingest a journalist's last 6–12 months of published work, summarize core themes, preferred story formats, recurring angles, and trigger topics.
Hook generation: Produce 5–10 candidate opening lines per journalist, each tied to a specific recent article or documented interest.
Angle matching: Score which aspects of your client's story best map to each journalist's demonstrated interests.
First-draft generation: Write complete pitch drafts at 70–80% quality—good enough to edit, not good enough to send blind.
List segmentation and prioritization: Rank your media list by relevance score so humans spend their review time on the highest-value targets first.
Subject line optimization: Generate and score multiple subject line variants for open-rate potential.
Follow-up cadence suggestions: Based on outlet type, beat velocity, and general response patterns.

Requires human judgment (don't automate this):

Relevance gating: Deciding whether the story is genuinely newsworthy for this specific person right now. AI consistently overestimates fit.
Relationship context: Knowing a journalist hates a certain angle, has history with your client, or is currently buried under a major breaking story.
Final voice and authenticity: Preventing pitches from sounding robotic. Journalists are very good at detecting AI-generated text in 2026.
Strategic angle crafting: The deepest creative connections—"This actually challenges the narrative you've been building for two years"—require human insight.
Ethical judgment: Avoiding exaggeration, manipulative framing, or claims your client can't back up.

Step-by-Step: Building the PR Pitch Agent on OpenClaw

Here's how to wire this together using OpenClaw as your agent platform. The architecture is straightforward: data in, research processing, segmentation, draft generation, human review queue.

Step 1: Define Your Agent's Data Inputs

Your agent needs three categories of input:

Client brief data: The news you're pitching, key messages, supporting data points, approved quotes, embargo details, target verticals.

Media list (raw): Journalist names, outlets, beats, email addresses. Export this from Muck Rack, Cision, or your existing spreadsheet. CSV format works fine.

Journalist research corpus: This is where OpenClaw earns its keep. Configure your agent to pull recent articles for each journalist on the list. You can connect to RSS feeds, outlet archives, or use web search APIs to retrieve the last 10–20 published pieces per contact.

In OpenClaw, you'd set up the agent's input schema to accept the client brief as structured text and the media list as a data source the agent iterates over. Each journalist becomes a "task" the agent processes independently.

Step 2: Build the Research and Profiling Module

This is the core of the agent. For each journalist on the list, the agent:

Retrieves and reads their recent articles (last 6 months, prioritizing the most recent).
Extracts key themes: What topics do they return to? What framing do they prefer (skeptical, enthusiastic, investigative, feature-style)?
Identifies specific hooks: Quotes, data points, or arguments from their work that connect to your client's story.
Flags relevance signals: Did they recently cover your industry? A competitor? A related trend?
Flags risk signals: Have they written negatively about your client's category? Are they currently covering a major unrelated story that might make your pitch poorly timed?

The output for each journalist is a structured research brief:

Journalist: Sarah Chen
Outlet: TechCrunch
Beat: Enterprise SaaS, security
Recent Focus (last 3 months): Zero-trust architecture adoption,
  AI in SOC operations, Series B+ funding rounds in cybersecurity
Preferred Format: News analysis (not straight news), 800-1200 words
Tone: Skeptical of vendor claims, responds to data
Relevant Hook: March 15 article on "Why Zero-Trust Promises Keep Falling Short"
  — argued that most vendors lack supply chain visibility.
  Client's product directly addresses this gap.
Connection Strength: HIGH
Risk Flags: None identified
Suggested Angle: Supply chain visibility as the missing piece in
  zero-trust — with client's pilot data as proof point

This research brief is what used to take 20–45 minutes per journalist. The agent produces it in seconds.

Step 3: Segment and Prioritize the List

Once every journalist has a research brief, the agent segments your list automatically:

Tier 1 (High relevance, strong hook): These journalists have recently covered your exact topic, and the agent found a specific connection point. Human review and final customization here is highest priority.

Tier 2 (Moderate relevance, adjacent coverage): They cover the broader space but haven't written about your specific angle recently. Still worth pitching but with a different framing.

Tier 3 (Low relevance or stale data): The agent couldn't find recent coverage that connects, or the journalist appears to have shifted beats. Flag for human decision: pitch with a broader angle, or remove from list.

Flagged for removal: Contact data appears stale (no articles published in 3+ months at listed outlet), beat mismatch, or risk signals identified.

This segmentation replaces the manual spreadsheet triage that typically takes 2–4 hours.

Step 4: Generate Draft Pitches

For Tier 1 and Tier 2 journalists, the agent generates complete draft pitches. Each draft includes:

Three subject line options (ranked by specificity and relevance to the journalist's recent work)
A personalized opening paragraph that references a specific article or theme from the journalist's recent output
The core pitch (adapted from the client brief, with the angle adjusted to match what the journalist cares about)
A suggested data point or exclusive offer relevant to this journalist's coverage style
A brief, non-pushy closing

Here's what a Tier 1 draft output might look like:

Subject Line Options:
1. "The supply chain blind spot in zero-trust (with data)"
2. "Re: your March piece on zero-trust gaps — we have pilot numbers"
3. "Zero-trust's missing layer: 14-company pilot results"

Opening:
Your March analysis of why zero-trust implementations keep falling short
nailed something most vendors won't admit — that network-level controls
mean nothing if you can't see what's happening in your supply chain.
We just completed a 14-company pilot that quantifies exactly how big
that blind spot is, and the numbers are worse than your piece suggested.

[Core pitch adapted to journalist's angle...]

Closing:
Happy to share the full pilot data under embargo, or connect you with
two of the CISOs who participated. No rush on timing — just wanted
this on your radar given your ongoing coverage.

The agent generates this for every Tier 1 and Tier 2 journalist. A human then reviews, edits, and approves before anything gets sent.

Step 5: Set Up the Human Review Queue

This is critical. Do not send agent-generated pitches without human review. The fastest way to burn your media relationships is to blast AI-generated pitches that contain hallucinated article references, outdated hooks, or that uncanny-valley tone that journalists have learned to spot instantly.

In OpenClaw, configure the output as a review queue—each draft presented alongside the research brief so the reviewer can quickly verify accuracy, adjust tone, and add any relationship context the agent doesn't have access to.

A senior PR person can review and approve an agent-prepared pitch in 3–7 minutes versus writing one from scratch in 30–90 minutes. That's where the real time savings compound.

Step 6: Iterate and Improve

After each campaign, feed response data back into your OpenClaw agent. Which pitches got responses? Which subject lines performed? Which angles fell flat? Over time, the agent's relevance scoring and angle matching improve based on actual outcomes rather than assumptions.

You can find pre-built agent components and workflow templates for media outreach on Claw Mart, including research modules, segmentation logic, and pitch drafting frameworks that other PR teams have already refined. Starting from a tested template and customizing beats starting from scratch every time.

What Still Needs a Human (Non-Negotiable)

To be direct: if you automate the wrong parts, you'll generate more pitches faster and damage more relationships faster. Here's what stays human:

The "should we pitch this person at all" decision. AI will always find some tenuous connection. A human needs to decide if the connection is real and if the timing is right.

Relationship memory. You know that this journalist ghosted you last time because your CEO said something tone-deaf in a podcast. The agent doesn't know that.

Final voice editing. Read every draft out loud. If it sounds like a machine wrote it, it'll read like a machine wrote it. Add a phrase that's genuinely yours. Remove the parts that feel templated.

Strategic creativity. The best pitches don't just match existing interests—they reframe them. "Here's something that contradicts everything you've been arguing, and here's why you should care" is a human insight.

Ethical review. Don't let an agent exaggerate your client's claims. Don't let it imply exclusivity you're not offering. Don't let it promise access you can't deliver.

Expected Time and Cost Savings

Based on the workflow above and consistent with what early adopters have reported publicly:

Task	Manual Time (75 journalists)	With OpenClaw Agent	Savings
Media list building + cleanup	4–8 hours	1–2 hours (human verification)	~70%
Deep research per contact	20–55 hours	1–3 hours (agent processing + spot checks)	~90%
Angle development	12–25 hours	Included in agent output + 3–5 hours human refinement	~75%
Pitch writing	18–37 hours	5–9 hours (human review and editing of agent drafts)	~70%
Segmentation and prioritization	2–4 hours	15–30 minutes + human review	~85%
Total	56–129 hours	10–20 hours	~75–85%

That's not a small improvement. That's the difference between one person spending three weeks on a single campaign and spending three to four days—with equal or better response rates, because the personalization quality is actually higher when the human reviewer isn't exhausted from doing the research manually.

One agency that tested a similar workflow (publicly discussed on PR podcasts in 2026) cut per-pitch research time from 45 minutes to about 12 minutes while maintaining their ~30% response rate. Their throughput tripled. A consumer brand using a comparable setup reported cutting total campaign time by 55% while increasing positive responses.

The cost math is equally clear. If a senior PR strategist's time costs $75/hour fully loaded, saving 50–100 hours per campaign saves $3,750–$7,500 per campaign. For an agency running 8–12 campaigns per quarter, that's $30,000–$90,000 annually in recovered capacity—capacity that can go toward the strategic and relationship work that actually differentiates great PR.

Next Steps

If you're running PR campaigns and spending more than 20 hours on research and personalization per campaign, this workflow will pay for itself immediately. Here's how to start:

Pick one upcoming campaign as your test case. Don't try to rebuild your entire PR operation at once.
Export your media list into a clean CSV with journalist names, outlets, beats, and any existing notes.
Build your agent in OpenClaw following the architecture above: research module → segmentation → draft generation → human review queue.
Browse Claw Mart for pre-built components. Other PR teams have already published research aggregation modules and pitch drafting templates that you can adapt rather than building from zero.
Run the agent alongside your manual process for the first campaign. Compare the agent's research briefs and drafts against what your team produces manually. Calibrate.
After one campaign cycle, feed response data back in and adjust relevance scoring.

If you've built a PR workflow or media research agent on OpenClaw that's working well, consider listing your components on Claw Mart through Clawsourcing. The PR community is actively looking for tested, proven agent templates—and you can monetize the work you've already done building yours. Learn more about Clawsourcing here.