I Manage 11 Client Accounts with AI Agents — Here's the Tech Stack | Social Ads AI

Running an agency with 11 active client accounts used to require a team of 5-8 people. Account managers, media buyers, reporting analysts, copywriters, project coordinators.

Today I run it with a team of AI agents and a small human crew focused on strategy and client relationships. Here's the complete tech stack that makes it possible.

The Agent Architecture

I run three primary AI agents, each with a distinct role:

Claude Code (Primary Brain)

Claude Code is the central nervous system. It handles complex analysis, strategic decisions, code generation, and orchestration of other agents. When a client needs a new landing page built, a campaign restructured, or a performance issue diagnosed, Claude Code handles it.

Sal (Discord Interface)

Sal bridges my mobile workflow to Claude Code's capabilities. Through Discord, I can issue commands, get reports, approve campaigns, and manage client work from my phone. Sal handles the translation between conversational requests and technical execution.

Sammy (VPS Workhorse)

Sammy runs on a dedicated VPS and handles the heavy lifting: video production, research pipelines, file processing, and Google Workspace operations. When bulk content needs to be generated or large data sets need processing, Sammy handles it without tying up my local machine.

The Tool Stack

Campaign Management

• Meta Ads API (via Pipeboard MCP) — Direct API access to all client ad accounts
• GoHighLevel — CRM, funnels, automation, client communication
• Google Ads — Search campaign management

Communication

• Discord — Internal team coordination and mobile command center
• Upstash Redis — Real-time cross-agent message passing and shared state

Content & Creative

• AI Image Generation — Ad creatives, social content
• Video Pipeline — Kling/SORA for video ad production
• Copywriting Framework — 333 frameworks from 22 marketing books, searchable by funnel stage and awareness level

Automation

• n8n — Webhook-driven workflows for cross-platform automation
• Playwright — Browser automation for tasks that require web interaction
• Google Workspace — Shared Drive for cross-agent file access

Daily Operations

A typical day looks like this:

1. 6:00 AM: AI agents run overnight performance checks on all 11 accounts
2. 7:00 AM: I review flagged issues and priority alerts on my phone via Discord
3. 8:00 AM: Strategy session — I make decisions, agents execute
4. Throughout the day: Agents handle monitoring, reporting, creative testing, and client communication
5. End of day: Summary report of all actions taken, results achieved, and issues flagged

Related Reading

• AI Agents for Business Hub
• AI-Powered Lead Generation
• Sales System Automation