Bringing daily local news back to 20 New Jersey towns
Patch asked me to lead a pilot in 20 markets that had no staff reporter. I designed an AI-assisted workflow, wrote the rules and kept an editor in charge of every story. Within a month, readers were clicking more. Then the model spread to other states.
The problem
Patch runs local news sites across New Jersey, but not every town has a reporter. In these 20 markets, readers were getting very little new local news, and the newsletters showed it.
The news itself still existed. Town halls, police departments and school districts put out releases, statements and meeting documents every week. What was missing was the time to turn that material into stories.
What I built
I used AI for the planning and the repetitive work, and left accuracy, judgment and voice to an editor.
-
01
Map every town
In ChatGPT Business, I built a beat map for each market: the offices, agencies and schools that publish news, and what readers there care about.
-
02
Plan the day
The beat maps fed a daily news roundup plan, so each morning started with a clear list of what was new in each town and what was worth covering.
-
03
Check the style
I used Claude to review drafts against AP and Patch style, which caught small inconsistencies before they reached an editor.
-
04
Edit and fact-check by hand
A person checked every name, number and date against the original source, cut anything the source didn't support, and kept Patch's voice.
-
05
Publish and measure
Stories went out on the sites and in daily newsletters, and I tracked what readers opened and clicked in Chartbeat and GA4.
- No editorializing.
- Nothing the source doesn't support.
- Specific dates instead of words like "yesterday."
What happened
Within a month, the newsletter click rate went from 4.22% to 6.07%, and the 20 markets added 795 net subscribers. Patch then rolled the workflow out to several other states.
I trained a newer colleague to run new New Jersey markets on the same model, and subscribers there grew 140%. On my own sites, a similar workflow saves me at least two hours a day across the eight to 10 stories I publish daily.
What made it work
The editor stayed in charge
AI saved time on planning and checking. People made the calls.
Official sources only
Working from public documents kept every fact checkable.
Rules on paper
Clear, written guidelines made the output consistent and the system easy to hand to someone else. That's why it could scale.
More case studies
See all work →Building something like this on your team?
I'd be glad to talk about content strategy, editorial operations and how AI fits into both.