Scope drift
Local, national, trade, and syndicated outlets were being mixed together.
AI WORKFLOW · BUILT + IN USE
A media list that
verifies itself.
Turning a high-stakes, hours-long PR research process into one reusable command—without automating away the professional judgment.
- For
- Independent PR consultancy
- Industry
- Hospitality + attractions
- System
- Claude skill + browser workflow
- Deliverables
- Google Sheet + evidence log
01 / THE FRICTION
The list wasn’t the hard part.
Trusting it was.
Building a client-ready media list meant moving between a brief, Muck Rack, live publication pages, journalist profiles, open-web research, and a shared spreadsheet. Every contact needed to be current, relevant, reachable, and backed by something they had actually written.
The first run exposed the real failure points: scope changed midstream, “national” meant different things to different people, some database records were stale or wrong, emails came from sources with different levels of confidence, and spreadsheet writes were fragile.
So the job wasn’t “make Claude search for reporters.” The job was to turn professional standards into a system that could enforce them every time.
02 / BEFORE THE BUILD
We did the work once.
Then audited everything.
The initial project was documented as it happened: the prompts, the decisions, the corrections, and the places where the tools got it wrong. That record became the blueprint.
Stale records
Database results included an incorrect byline and a reporter who had left.
Missing proof
The verification standard existed in working memory, not as a repeatable checklist.
Fragile handoff
Spreadsheet pastes could land in the wrong place when the browser layout changed.
03 / THE SYSTEM
One command.
A controlled workflow.
The finished skill does not jump straight into research. It collects the expensive-to-get-wrong decisions first, confirms the brief, then runs a documented sequence with evidence attached.
- 01Define
Client, subject, geography, date range, comparables, destination.
- 02Confirm
Select each outlet tier and approve a five-line brief before research begins.
- 03Discover
Search the media database, live publications, and relevant peer coverage.
- 04Verify
Check every byline, article, date, current role, and contact source.
- 05Deliver
Write the sheet, preserve the evidence log, and report what was excluded.
04 / THE OUTPUT
What comes out
the other end.
A Google Sheet, which is what the client wanted. The list of names is the easy half; the last column—where the sources disagreed and what the system did about it—is the half that takes judgment.
The brief: reporters who have covered new resort and attraction openings in the last twelve months. Regional trade and national consumer travel. No syndicated wire pickups.
| Name | Outlet | Tier | Email source | Relevant coverage | Status | Verification notes | ||
|---|---|---|---|---|---|---|---|---|
| 1 | Marisa Ocampo | Coastal Hospitality Review | Regional trade | m.ocampo@coastalhospitalityreview.example | Outlet masthead | 2026-06-18 2026-03-04 | Verified | Database byline reads “Marisa T. Ocampo.” Last six published pieces run as “Marisa Ocampo.” Using the published spelling. |
| 2 | Dev Raghunathan | The Getaway Desk | National consumer | dev.r@getawaydesk.example | Outlet contact page, updated Jul 2026 | 2026-07-29 2026-02-11 | Verified | Database lists an older address that no longer appears anywhere on the outlet. Using the masthead address. Both kept in the evidence log. |
| 3 | Priya Anand | Meridian Travel Weekly | Regional (reclassified) | — | None verifiable during the run | 2026-08-02 2026-05-15 | Verified | Tagged national in the database; last eleven pieces are regional. Reclassified, left in scope. Email left blank rather than guessed. |
| 4 | Tom Bexley | Harborline Monthly (former) | Regional trade | — | Not applicable | 2025-10-09 | Excluded | Left the outlet in January 2026, no byline since. Top-ranked database result. Kept on the sheet with its reason rather than dropped. |
The real sheet lands in the client’s own drive, sectioned by outlet tier, with the evidence log on a second tab. The contacts above are invented for publication; the columns, and the kind of note the system writes into the last one, are what the workflow actually produces.
05 / THE GUARDRAILS
The mistakes became part of the product.
01
No silent assumptions. The workflow asks which outlet tiers matter every single run.
02
No contact without proof. Every usable record needs a byline, article, date, URL, and current-role check.
03
No invented fields. If an email or date cannot be verified from a page opened during the run, it stays blank.
04
No hidden exclusions. Stale, wrong, out-of-scope, and unverified results are retained with a reason.
05
No blind spreadsheet writes. Exact ranges are selected and checked after every paste.
06 / WHAT CHANGED
From rebuilding the job
to directing the system.
BEFORE
Re-explain the process
Search across several sources
Manually prove each contact
Reformat and re-section the sheet
Remember every quality check
NOW
Run /media-list
Answer a focused intake
Confirm the outlet tiers
Review flagged or uncertain records
Use the verified list and pitch angle
THE PRINCIPLE
The professional still directs the work. The system handles the repetition—and shows its receipts.
07 / THE HUMAN PART
Automation with an adult in the room.
The PR professional chooses the client strategy, geography, comparables, and outlet tiers. She can redirect the research as the story takes shape. The workflow handles the procedural load, surfaces uncertainty, and preserves the evidence needed for a confident final review.
This case study is published with identifying client and contact information removed. The underlying workflow was built from a live project and is currently in use.
YOUR PROCESS DOESN’T HAVE TO BE PR
What do you rebuild
every time?
Show me the process If a valuable job depends on the same research, decisions, and handoffs every time, there may be a system hiding inside it.