Case Study

How Ziv Health manages 11 AI agents from Slack

One operator. One Slack channel. A regulated telehealth business with 139 scheduled jobs, 150+ event types, and 11 autonomous agents — all managed by voice and text from a phone.

Powered by Voice for Slack from Shulam.

11
AI Agents
Named souls with AAIN identifiers
139
Cron Jobs
Scheduled phases running 24/7
150+
Event Types
Patient lifecycle, revenue, compliance
1
Operator
Managing everything from Slack
<5s
Alert Latency
Payment → Slack notification
10
Registered Actions
Available via natural language

The problem

Ziv Healthis a telehealth platform that prescribes GLP-1 weight loss medications, men's health treatments, women's hormone therapy, and longevity compounds across all 50 states. It runs on Shulam with 11 named AI agents managing patient intake, prescription routing, pharmacy fulfillment, compliance monitoring, SEO, ad optimization, retention, and revenue intelligence.

The operator needed to manage this entire system while mobile — walking between meetings, commuting, or working from a phone. Opening a laptop to check a dashboard for every alert, every payment, every churn signal was not viable. The agents were autonomous, but the operator still needed real-time visibility and the ability to intervene.

The requirement: manage 11 agents, 139 scheduled jobs, and a regulated healthcare business from a single Slack channel, using voice memos and natural language text — no dashboard, no laptop, no context switching.

The solution

Voice for Slack was built in a single session and deployed to production the same day. Four components were wired:

E-ZIV-OPS-01: Operations Channel

SlackOpsService — incoming webhook for critical alerts. Dedup window prevents alert storms. Urgency-based formatting (critical/high/medium/info). Every alert includes revenue impact in dollars, recommended action, and admin deep-link.

E-ZIV-OPS-03: Event Wiring

Five EventBroker subscriptions route real-time events to Slack: payment received, new patient signup, churn risk detected, subscription cancelled, Stripe dispute created. Three scheduled phases deliver daily revenue digest (6pm ET), refill due digest (8am ET), and weekly G-Stack scorecard (Monday 7am ET).

E-ZIV-OPS-04: Agent Dispatch

OperatorDispatchService with 10 registered actions. Two-tier NLP classification (pattern match + Claude Haiku). Destructive actions (pause ads, process refunds) require confirmation within a 5-minute window. Every dispatch is logged in the HIPAA audit trail.

Voice Transcription

Whisper-based speech-to-text converts voice memos to text. The transcript passes through the same two-tier NLP pipeline as typed messages. Voice and text are functionally equivalent — the operator switches between them based on context.

Implementation timeline

Day 1

Slack channel created

#zivhealth channel in the Shulam workspace. Incoming webhook configured. First test message received.

Day 1

Agent alerts wired

SlackOpsService connected to EventBroker. Payment notifications, churn alerts, dispute warnings, MDI case updates — all posting in real-time.

Day 1

Scheduled digests configured

Daily revenue digest at 6pm ET. Refill due digest at 8am ET. Weekly G-Stack scorecard every Monday at 7am ET. Morning brief already wired.

Day 1

Agent dispatch layer deployed

OperatorDispatchService with 10 registered actions. Natural language classification via two-tier NLP. Destructive action confirmation guard. Full audit trail.

Day 2

Voice transcription activated

Whisper-based voice memo transcription. Record a voice note in Slack → transcribed → classified → dispatched to the right soul → result posted in thread.

Results

Operator manages 11 agents, 139 cron jobs, and a regulated telehealth business from one Slack channel

Every payment notification arrives in Slack within 5 seconds of Stripe confirmation

Churn risk alerts include the patient's negative sampling cohort and estimated MRR loss

Destructive actions (pause ads, refunds) require explicit confirmation — no accidental commands

Daily revenue digest at 6pm ET replaces end-of-day dashboard check

Weekly G-Stack scorecard at 7am ET Monday replaces the weekly review meeting

Morning brief at 7am ET daily — one scroll covers the entire business

Zero PHI in Slack — all messages use anonymized patient IDs, never names or emails

“I manage a regulated telehealth business with 11 AI agents from my phone. Voice for Slack is how autonomous operations actually works in practice.”

— JJ, Founder & CEO, Ziv Health

Want Voice for Slack for your business?

Every Shulam Lighthouse client gets Voice for Slack on Day 1 — no additional setup, no additional cost.