The claude system that runs our daily outbound ops

Hey there,

David here.

There is a part of running outbound that rarely gets talked about. The daily tweaking.

Every day, across every live campaign, there are decisions waiting. Which campaigns to scale. Which to stop. Which inboxes are quietly hurting deliverability. Which campaigns are about to run dry. Leave those alone for a week and results drift.

We used to work through all of it by hand. It took hours. So we built it into Claude instead, and I recorded the full walkthrough.

Watch: The Claude System That Runs Our Outbound Every Day

Below is the logic underneath it, in case you want to build your own version.

01The three routines

• Collect. Pulls analytics from the email sequencers, from LinkedIn, and from every mailbox in the sending infrastructure, then writes all of it to Supabase.
• Report. Posts that data into five Slack channels, one per question, with traffic lights so each read takes seconds.
• Act. Reads our replies in those channels and performs the work.

The first two are reporting. The third is the one that removes the hours.

02The five questions, and the rules behind them

1. Are we actually on track for this client?

A score against a lead target. The target is not a round number someone picked. It comes from that client's average deal value, close rate and sales cycle, worked backwards into a cost per lead and the number of leads that make the program profitable. If the lead volume does not clear that bar, the program is not working, however good the reply rate looks in isolation.

The score also weighs how long we have been working together. Message-market fit usually takes a few months to find. Before it lands, output is noisy. After it lands, positives climb quickly and scaling gets much easier. Holding month one to month six's standard tells you nothing useful.

We read it across both channels, over 30 days and over 7 days, next to send volume.

2. Stop, leave, or scale?

Our bar on email is roughly one positive reply per 50 sends. Clear that consistently and the campaign earns more volume. The same ratio on LinkedIn is only average, because the volumes and the medium are different, so the two channels get judged on separate scales. A campaign sitting below the bar with no trend upward comes out rather than limping along.

3. How much runway is left?

Leads remaining, and more usefully, how many days that buys at the current send rate. Runway is the number that matters. A campaign with 500 leads sending 200 a day is nearly dead, and the raw lead count hides that. Red means top it up today, not this week.

4. Is deliverability holding?

Two checks per inbox. The first is the warm-up score, and under 97 gets flagged. The second is how many replies that domain has actually delivered, because a sender can look healthy on paper and still be landing nowhere. Either flag and the inbox rotates out, goes back onto warm-up, and can return later if it recovers. Rotating early costs far less than burning a domain.

5. Did anything disconnect?

Mailboxes drop their connection on their own. Nobody notices for days unless something is watching every morning, and a disconnected sender is a campaign quietly sending less than you think'

The channels stay read-only until someone replies.

When an inbox should come out, I reply in the thread. Something like "rotate this sender out, the warm-up has tanked." The third routine picks that up and performs the rotation. Same for topping up leads. Same for pausing a campaign.

The system surfaces the decision, a human makes the call, and the system executes it. Nobody opens four tools to action a two-word instruction.

03If you want to build this

Start with the reporting, not the automation. Pick the five questions, get the data into one place once a day, and post it somewhere you already look. A manual version beats checking dashboards when you happen to remember. Add the execution layer only once you trust what the reporting is telling you.

For our clients this runs in the background, and most of them never see these channels. That is the point. The complexity stays on our side.

04One small thing you can do this week

Before you launch your next campaign, write down the number that will kill it, and the date you will check that number. Put it in the campaign doc, next to the copy.

Kill decisions get delayed when nobody agreed the bar in advance. A campaign that is clearly underperforming gets one more week, three times over, because the conversation starts only once someone is already attached to it. Deciding the bar before you have any feelings about the campaign is most of the discipline.

Ours is one positive per 50 sends on email, checked at a fair sample. Yours will differ. Writing it down beforehand is the part that matters.

I am curious which of the five you are answering properly today, and which one you are guessing at. Hit reply and tell me, I read every response.

PS. If you want us to build a similar GTM system for you, grab 20 minutes here.