Most founders meet AI the way they’d meet one talented freelancer: a chat here, a task there, impressive moments, no compounding. That’s not how you run a company. You run a company with a team — people with roles, standing responsibilities, and a calendar. The unlock isn’t another skill file. It’s realizing that a skill + access + a goal + a schedule = an employee, and that founders need a whole operating team of them working around you while you sleep. This page is the definitive, continuously updated map of that team: the roles, the jobs each role runs, paste-ready prompts, and live receipts from our own fleet.
(Last week the newest hire on our team was the podcast citation publicist. Next week it will be something else. The roles below are the durable structure — the individual jobs keep growing.)
Your AI operating team — six roles every founder needs
| Role | Human-company equivalent | What they run |
|---|---|---|
| Chief of Staff | keeps your day pointed at the goal | morning brief, priority triage |
| Publicist | collects and amplifies your proof | proof harvest, citation sync, podcast inventory |
| Security Guard | watches the doors at night | malware/admin/registrar watch, uptime, disavows |
| Analyst | tells you what the numbers mean | weekly MAA reports, SEO deltas, GSC truth-checks |
| Content Editor | turns raw material into published work | article waves, repurposing, interlinking |
| Librarian | makes the whole team smarter every week | skills audit, SOP updates, schema sweeps |

How to use this page: install skills and grant access first (the full walkthrough is the agent activation checklist), then hire one role a week, in the order below. Tell Claude: “Create a scheduled task” and paste the prompt seed. Two employees beat ten job descriptions.
The first five hires (in order)
1. Chief of Staff — the morning brief (daily, weekdays)
“Every weekday at 6:30am, brief me: today’s calendar, emails needing replies, yesterday’s wins from my reports, and the ONE thing that moves my goal forward. Keep it under a minute to read.”
Access: Gmail, Calendar. Why first: it proves the loop works and you feel it every morning.
2. Publicist — the weekly proof harvest (Fridays)
“Every Friday, run my positive-mentions-harvester skill on the week’s new mentions — podcasts, reviews, shoutouts, client results. Score each on the 30-point scale, add winners to my proof library, and flag anything worth a Dollar-a-Day boost.”
Access: web search; your proof sheet. Output: a proof library that compounds instead of rots.
3. Publicist — the authority-citation sync (Mondays)
“Every Monday: check my Podchaser credit count and Listen Notes results. Triage new suggested credits individually (reject mere mentions and name collisions). Update my /podcast-appearances/ page counts and dated update block. Verify my skill downloads and key links return 200. Log my pages’ AI-citation counts. Report the deltas.”
Access: WordPress app password, analytics. This is the job that keeps your citations current — method in the podcast inventory guide and the complete platform map.
4. Security Guard — the SEO + security watch (weekly)
“Weekly: scan my email for security alerts (malware scans, new admin accounts, registrar changes) and flag anything abnormal. Pull new referring domains for my site; list new toxic/spam domains in disavow format. GREEN/AMBER/RED summary. Read-only — change nothing.”
Access: Gmail, Ahrefs or Search Console. Why: our version caught a live spam-cloak and an injected-admin pattern that manual checking missed for days.
5. Librarian — the weekly skills audit (Fridays)
“Every Friday, review what my agents produced this week vs. what my skills say should happen. List: resolved and verified, still open, and the one blocker that needs ME. Update the relevant skill file with anything we learned.”
Access: your task outputs. Why: this is the compounding loop — the employee who makes every other employee smarter.

The full catalog — what our own fleet runs
Generalized from the real scheduled jobs running on our accounts right now (31 active on one machine plus cloud tasks, before counting client-specific ones):
| Cadence | Job | Role | What it does |
|---|---|---|---|
| Daily | Morning brief | Chief of Staff | Calendar + inbox + one priority |
| Daily | Article wave | Editor | Transcripts → staggered posts (Content Factory) |
| Daily | Intake watcher | Chief of Staff | Watches request queues; stages provisioning |
| Weekly | Authority-citation sync | Publicist | Podcast counts → pages → AI-citation log |
| Weekly | Proof harvest | Publicist | New mentions scored into the proof library |
| Weekly | SEO + security watch | Security Guard | Malware/admin/registrar alerts + disavows |
| Weekly | Uptime + interlinking | Security / Editor | Probes the site fleet; adds hub cross-links |
| Weekly | Skills audit | Librarian | Resolved vs open; updates the skill files |
| Weekly | Client MAA reports | Analyst | Per-client measurement & analytics drafts |
| Monthly | Full podcast re-inventory | Publicist | 5-platform sweep, diff vs master sheet |
| Monthly | Schema sweep | Librarian | Person/Article schema QA across every site |
| Monthly | SEO delta re-audits | Analyst | Rankings/backlinks movement per property |
| Monthly | Grokipedia/Wikidata readiness | Publicist | Entity records current, corrections staged |
| Monthly | Agent library refresh | Librarian | Skill packs + download pages kept current |
Live receipts (updated by the jobs themselves)
As of August 3, 2026: our security watch caught two hidden P1 incidents in its first week, including a live spam-cloak serving 3,000 junk URLs · the citation system holds 619 verified Podchaser credits across 353 podcasts · Ahrefs Brand Radar counts 67 citations of blitzmetrics.com inside AI answers (Grok: 17, Google AI Overviews: 15, Gemini: 13) — earned while classic organic metrics on the same pages still read near zero (full analysis: why Ahrefs data is superficial for citation work) · a client site got 51 verified fixes applied by one overnight job. This block is refreshed by the Monday sync job — the page you’re reading is maintained by the system it describes.
Humans in the loop — we implement this with people, in person

None of this is AI-generated slop shipped from a distance. We fly to the rooms: the DSDT AI Summit in Detroit, Digital Wichita at Wichita State University, DealCon in Denver, JVA Align in Dallas and Chicago — sitting next to founders while their first scheduled job fires. The agents do the recurring work; humans teach humans, check the receipts, and decide what matters. That’s the operating team: your judgment at the top, six tireless employees underneath, and a community of builders comparing notes.

Community examples
Sigrun’s mastermind members run a daily security monitor (it caught a live casino-spam injection on a member site) and a quarterly software-subscription audit that started as one member’s ~$4,000/yr savings run and became a shared skill. Jack Wendt’s team runs six scheduled tasks in his Claude Mainframe, including nightly file organization and SOP retrospectives. A family-law client gets weekly leaderboard check-ins without anyone remembering to run them. Running your own? Send us your job list and receipts — the best ones get featured here with a link to you.
Hire your first employee in the next ten minutes
- Install the skills: dennisyu.com/dealcon
- Follow the activation checklist — Claude configured, access granted, goals clarified
- Paste hire #1’s prompt seed above and say: “set this up as a scheduled task.” Add one role a week.
Related: how an agent scheduled its own dashboard · keeping your Mac running 24/7 for agents · the Claude Max plan math · the Task Library

