Five Agents. Zero Salaries.

The operator's job isn't to do the work. It's to walk the row, coffee in hand, and decide what to do with the work that's already done.

5 + 1
Agents plus concierge
1

The Five Desks

Desk 1

Copy Drafter

CadenceOn demand (new positions, new experiments)

InputPartner content (transcripts, blog posts, about page)

OutputLanding page drafts, email sequence drafts, experiment variants

Time saved3-4 hours → 45 minutes per position

Replaces the blank screen, not your judgment. 70% of the way there. You do the remaining 30%.

Desk 2

Experiment Analyst

CadenceDaily

InputAnalytics data, experiment log

OutputPer-experiment summary: sample size, lift, confidence interval, recommendation (call/kill/run)

Anomaly detectionPosition 3, CTR dropped from 4.2% to 1.1%. Possible broken redirect.

The anomaly flag alone is worth the entire agent.

Desk 3

List Health Monitor

CadenceWeekly

InputEmail engagement data (opens, clicks, bounces, unsubscribes)

OutputEngagement tiers (active/cooling/dormant/dead), deliverability signals, recommended actions

Catches drift at week 2, not month 3. Fires reactivation sequences automatically.

Desk 4

CCI Pattern Scanner

CadenceMonthly

InputCross-network commerce intelligence database

Output3-5 patterns with recommended actions (cross-promotions, product categories, timing)

At 2-3 positions: not much. At 5+: "Woodworking + home renovation subscribers purchase tools at 2.3x rate."

Desk 5

Daily Digest Compiler

CadenceDaily at 6 AM

InputOutputs from all four other agents + revenue data

Output5-line morning summary

Example"Revenue: $347 (up 4% WoW) | Experiments: 1 winner called | List: all green, +89 net | CCI: scan in 12 days | Anomalies: none"

15 seconds to read. The operator knows the state of everything.

2

The AI Concierge

The Sixth Role

Front porch, not bullpen

Chatbot embedded in toll position pages. Real-time revenue layer.

Context-aware help
Answers visitor intent before the email sequence begins.

Example conversation

"We have 3 kids and it's raining. What should we do?"
Context-aware recommendations with booking links.

Every conversation is data that feeds back into Copy Drafter, Experiment Analyst, and CCI Scanner.

Incremental layer

+8-15%

Incremental revenue on top of email sequences. Pennies per conversation.

The concierge earns by turning ambiguous visitor needs into specific partner clicks.

3

The Automation Hierarchy

Level 4: Never Automate

Decision-making. Which products. Which partners. Which experiments. Trust decisions.

Level 3: Experiment Monitoring

Recommendations, not actions. Tells you what to do. You decide.

Level 2: Draft Generation

Copy Drafter generates options. You choose. Nothing reaches subscribers without approval.

Level 1: Automate First

Data collection & reporting. Lowest risk. If it breaks, you miss a morning report.

Automate behind the trust layer. Keep the human on everything facing the trust layer.

4

The 30-Hour Budget

Activity Before (hrs/mo) After (hrs/mo) Savings
Data collection817 hrs
Copy drafting1037 hrs
Experiment monitoring615 hrs
Reporting40.53.5 hrs
Partner communication220 hrs
Total307.522.5 hrs

Effective hourly rate

$791/hr

At $5,938/month, the operator keeps the judgment work and gives away the clerical work.

5

What Not To Automate

×

Partner Relationships

AI drafts the outreach email. You show up on the call.

×

Deal Negotiations

Copy Drafter formats the memo. You sign it.

×

Ethical Product Curation

"Would I recommend this to a friend?" is a human question.

×

The Publish Button

Every email passes through human eyes. This is the feature, not the inefficiency.