On the morning of Monday, October 12, 2026, socalpitch.com published two posts about three hours apart: “SoCal Youth Soccer Standings — Week of October 12, 2026” and “Featured Club: Ojai AYSO Region 147, Ojai.” Both are still on the site’s news page. Agents did nearly all of the work on both. I did two things that morning: I cleared one verification check the agents are not allowed to touch, and I picked Ojai.
That split is the subject here. Episode 1 was the site. Episode 2 is the org chart.
The site runs on 20 scheduled jobs and 12 written playbooks
Every recurring task on SoCal Pitch is a scheduled job: a written instruction that fires on a calendar, does one job, and reports back. As of October 22, 2026, my scheduler runs 20 recurring jobs that exist only for this site. One more job checks search indexing for all of my sites, and a second agent, on a separate account, posts the Instagram queue.
Grouped by function, the way an org chart would group people:
| Function | Jobs | Cadence |
|---|---|---|
| Content | Standings roundup (nine leagues); Featured Club; Featured Club delivery check; monthly news scan; weekly content radar; Tuesday email draft; Instagram batch | Twice a week to monthly |
| Directory and data | Club pipeline (inside Featured Club); club legal-entity resolver; Top 50 index refresh; pre-season standings feed re-map; fall registration re-verify (through Oct 31) | Weekly to yearly |
| Quality | Outbound link sweep; stale-date sweep; rankings cross-check; external ranking source watch; monthly full site audit | Twice a month or monthly |
| Search | Position and query check; search health (shared with my other sites) | Twice a month, weekly |
| Business | Business-lane heartbeat; quarterly LinkedIn slate; trainer outreach follow-up | Monthly to quarterly |
Source: my live job scheduler, read Oct 22, 2026. The count excludes jobs I retired (below), one-time jobs, and the jobs shared with my other sites.
The jobs do the work. The playbooks tell them how. There are 12 that the SoCal Pitch jobs load, six written for this site and six shared with my other sites:
| Playbook | One-line job |
|---|---|
| Featured Club engine | Reconcile club submissions, build a 10-club candidate pool, stop for my pick, then build, publish, and archive |
| Spotlight archive wording | Turn every past spotlight’s “this week” language into dated past tense, so only the newest one speaks in the present |
| Directory integrity | Add or correct a club on strong evidence; never remove one without my approval |
| Business lane | Hold the publishing floor and the one-channel rule for the business-of-youth-soccer posts |
| Design compliance | Check every change against the design system before and after it ships |
| Site audit | The monthly full-site check |
| Publish gate | SEO checklist before publishing, an indexing request after |
| Render check | Label every finding by what a visitor actually sees, not what the source code says |
| Blocked-verdict rule | Before calling anything “blocked,” check whether it has worked before |
| Sitemap repair | Diagnose and fix sitemap problems before escalating them |
| Search console operations | One entry point for all indexing and coverage work |
| Transport choice | Decide whether a job writes over the server, a render check, or a browser |
I won’t publish the text of any playbook. What they do is the useful part, and the wording is mine.
The human sits at four checkpoints, not at the top
A normal org chart puts the owner at the top with everything flowing up. Mine doesn’t look like that. Most jobs never come to me. They run, verify their own output, and log it. The chart is wide and flat, and I sit at four points where a decision can’t be delegated:
- The pick. Every Monday the Featured Club job builds a candidate pool and stops. Nothing publishes until I choose. The choice is the editorial voice of the site, and it’s the one call I’d be least willing to hand off.
- The verification check. One league’s standings sit behind a human-verification check. The standings job collects the other leagues, then pauses for me to clear it. The agents are not allowed to solve it, and that rule isn’t negotiable. Once I clear it, the job finishes and publishes without asking again.
- The send. The Tuesday email is drafted Monday afternoon and stops as a draft. I read it Monday evening and schedule it. Nothing reaches a family’s inbox on an agent’s say-so.
- The removal. The directory job can add a club or fix a link on strong evidence. It can’t remove a club. A club that looks defunct gets flagged to me, because a wrong removal costs a real club real families.
Some approvals are narrower and written down in advance. The fall registration job, for example, has standing approval to move past-dated listings to “Already closed,” and nothing else. Anything outside that box is reported, not done.
Override is a job too
The part of the chart that doesn’t show up in a box is what I’ve cancelled. Three examples, all from the registry:
- Sept 2, 2026: I retired a weekly grant-scan job outright. It produced reading, not results.
- Sept 10, 2026: I merged the club pipeline job into the Featured Club job the same day it was created, so one job owns a club from submission to spotlight.
- Sept 15, 2026: I retired a media-query sweep mid-run.
And one where I overrode the plan rather than a job: on Oct 9, 2026, I parked the sponsorship program until traffic supports it. The agents had the pieces ready. Readiness wasn’t the question.
Running this well has been mostly about deleting things. An agent will happily keep doing a job nobody needs. Deciding what stops is the manager’s work, and it hasn’t moved.
The strongest objection: this is just automation with extra steps
A web professional could fairly say: cron jobs and checklists have existed for decades, and nothing here needed an agent. I’ll grant half of it. The scheduling is ordinary. What isn’t ordinary is what sits inside each job: reading an unfamiliar club website and deciding whether the club still exists, writing a standings roundup in plain English, catching that a club card points at the wrong organization even though the link loads fine. Those used to be a person’s judgment calls. Now they’re a job’s first draft and my last look.
Where the objection lands is cost of attention. Twenty jobs can fail in twenty ways. The org chart is only as good as the reports that come back from it, and that’s the subject of Episode 4.
The bottleneck is still the one person
The chart is only as true as my own follow-through at those four points. The pick has to land before the Featured Club deadline. The verification check has to be cleared the morning the standings run. The email has to be read Monday night. A flagged club has to get a decision. When any of those slips, the work stalls at my desk, and the chart describes a plan instead of an operation. Agents removed most of the work. They didn’t remove the manager.
A job count is not a quality score
Twenty jobs measures coverage, not quality. Some of them run twice a year. Several needed fixing more than once, and those fixes are the subject of Episode 4. The count comes from my own scheduler, so it rests on my word. The work itself is on socalpitch.com, where anyone can judge it. The hours behind it are in Episode 1: H2, an estimated 190 hours from March 15 to September 29, in a range of 155 to 215.
Next: Episode 3, the week of November 9. The week the automation couldn’t reach Drive.
Every episode, with the plan and the schedule, is on the No Agency Required series page, where you can also get each one by email.