Cutting editing time 40% at a 180K-subscriber newsletter
Client: Points Path, a daily travel and personal finance newsletter
Role: Managing editor
What I built: A custom AI editorial system covering QA, voice, and error tracking
The problem
Points Path publishes daily to a list of more than 180,000 subscribers, in a category where accuracy is not negotiable. A single outdated welcome bonus is the kind of error readers notice and remember. Not to mention affiliate partners, where real money is on the line.
The team needed to scale output without letting quality slip in the process. The usual answer is to hire another editor, but this is a lean team. The other answer, running drafts through a general-purpose AI tool, tends to produce copy that's confidently wrong about exactly the details that matter most.
What I built
I built a custom editorial system with four layers, designed around Points Path's own standards rather than generic best practices:
Voice and standards. The publication's style rules, tone, and structural conventions, encoded so they get applied the same way on every draft.
Quality checks. Verification steps for the things most likely to be wrong and most costly when they are: figures, terms, links, and claims that need sourcing.
Error tracking. A running log of every failure mode we've hit, including the system's own. Forty-two documented entries and counting. Each one gets written up and fed back in, so the same error doesn't ship twice.
Governance. Clear rules for where AI is allowed to operate and where a human editor has to sign off. Anything the system writes itself gets held to the same sourcing bar as a contributor draft. That rule exists because a tip it generated once carried an unverified hotel occupancy cap and nearly ran.
The system runs on contributor posts as part of daily production. I'm the one primarily using the tool, but it's documented and sharable so that it could easily be handed off to another colleague to operate.
The results
Editing time down roughly 40%, with no drop in quality
Open rate up from 49% to 55%, while the send list grew about 50% and send volume rose about 30%
Unsubscribe rate per send roughly halved over the same period, holding under 0.12%
Frequency scaled to daily sends, with a second afternoon send about twice a week
The system compounds, because each mistake because part of the playbook. Every error the system catches, and every one it misses, makes it harder for that error to happen again.
What this would look like for your team
The layers are the same, but the contents are completely yours: your voice, your standards, your failure modes. Most teams and brands start with a two-week audit of one workflow.