How to Set Up an AI Pipeline for Solo Agency Owners in 45 Minutes

Most solo agency owners use AI the same way: open ChatGPT, type a prompt, copy the output, paste it somewhere, fix the formatting, add context it did not have, and repeat. It works, sort of. But it is manual, inconsistent, and you are starting from zero every single time.

An AI pipeline is different. It is a structured, repeatable workflow where AI executes multiple steps in sequence using your actual business context. Instead of typing "write me a cold email," you run a pipeline that reads your prospect data, checks your previous outreach templates, scores the lead, and drafts an email that sounds like you wrote it. Then it saves the output, logs the run, and optionally writes the lead to your CRM.

This guide walks you through setting up real AI pipelines using GridWork HQ. Not theory. Actual steps, from installation to your first automated pipeline run.

What an AI Pipeline Actually Is

Let me be specific, because the term gets thrown around loosely.

An AI pipeline in this context is a Markdown definition file that describes a multi-step workflow. When you trigger it, the pipeline server reads the definition, spawns a Claude Code process with those instructions as the system prompt, and the AI executes each step in order. It can read files from your knowledge vault, call external tools (like Brave Search for prospecting), write output files, and push data to integrations like Notion.

The key differences from chat-based AI usage:

Context persistence. The AI reads your knowledge vault on every run. Your pricing, your templates, your client history, your evaluation criteria. It is not starting from scratch.

Structured output. Pipelines produce formatted deliverables, not conversational responses. A prospect pipeline outputs a scored lead table. An audit pipeline outputs a detailed technical report. A content pipeline outputs a month of blog posts with metadata.

Repeatability. Run the same pipeline on different inputs and you get consistent quality. The process is defined once and executed many times.

Automation. Pipelines can run on cron schedules without you triggering them. Weekly prospect sweeps, monthly scope audits, Friday client updates. Set it and the system handles it.

Why Solo Agency Owners Need This

You already know you are spread thin. Here is where pipelines pay for themselves specifically:

Prospecting takes hours you do not have. Finding leads, checking their websites, hunting for contact emails, scoring whether they are worth reaching out to. A prospect pipeline does this in minutes and outputs a ranked table with pitch angles.

Consistency drops when you are busy. When you have three active client projects, your outreach quality drops, your proposals get sloppy, and your content output falls off. Pipelines maintain quality regardless of how busy you are.

You cannot afford to hire yet. A virtual assistant costs $1,500-3,000/month. A junior developer costs more. AI pipelines handle the research, drafting, and operational tasks that would otherwise require another person.

Repetitive work kills momentum. Writing the same style of proposal, running the same type of site audit, generating the same monthly report format. These are exactly the tasks that structured automation handles best.

Setting Up GridWork HQ

GridWork HQ ships as a ZIP file containing two packages: a Next.js dashboard and a Node.js pipeline server. Here is the fast path to running.

Prerequisites (5 minutes)

You need three things installed:

• Node.js 20+ (22 recommended). Check with node -v.
• Bun (JavaScript runtime and package manager). Install with curl -fsSL https://bun.sh/install | bash.
• An Anthropic API key. Sign up at anthropic.com and generate a key. This is what powers the AI pipelines.

You also need a GitHub account for dashboard authentication.

Install and Configure (10 minutes)

Unzip the download and run the setup wizard:

unzip gridwork-hq-v*.zip
cd gridwork-hq
bun run setup

The setup wizard walks you through every configuration option: your API keys, GitHub OAuth credentials, Notion integration (optional), Telegram alerts (optional), and server settings. It writes a .env file for each package.

After the wizard completes:

# Terminal 1: Start the dashboard
cd gridwork-hq && bun install && bun dev

# Terminal 2: Start the pipeline server
cd pipeline-server && bun install && bun run build && bun start

Open http://localhost:3000, sign in with GitHub, and you are looking at Mission Control.

Running Your First Pipeline

The prospect pipeline is the best starting point. It requires zero setup beyond what you have already done and produces immediately useful output.

Trigger the Pipeline (2 minutes)

From Mission Control in the dashboard, find the Prospect card. Enter a query like:

prospect barbershops in Buckhead

The format is always prospect [business type] in [area]. Pick an industry and location you actually work in. The results will be more useful if they are real leads you might contact.

Watch It Work (5-10 minutes)

The pipeline server spawns a Claude process that:

1. Searches for businesses matching your query using Brave Search
2. Visits each website and checks mobile responsiveness, page speed, design quality, SEO basics, and broken links
3. Discovers contact emails by scraping contact pages, mailto links, and footer text, with a search fallback
4. Scores each lead as HOT (bad site, clear need), WARM (decent site, room to improve), or COLD (good site, low need)
5. Writes qualified leads to Notion if you configured the integration

You can watch the progress in real-time via SSE streaming in the dashboard. Each step logs as it completes.

Review the Output (3 minutes)

When the pipeline finishes, you get a ranked table:

#	Business	URL	Email	Score	Issues	Pitch Angle
1	Example Barber	example.com	info@example.com	HOT	No mobile, slow load	Full rebuild
2	Other Shop	other.com	NOT FOUND	WARM	Dated design, no SSL	Redesign + SEO

The output is saved to .output/prospect-buckhead-barbershops.md and cached for 7 days. Hot and warm leads with confirmed emails are written to your Notion leads database.

From here, you can run audit [business] on any hot lead for a detailed technical report you can send as a free value-add, or run outreach [business] to draft a personalized cold email.

Automating with Cron

Running pipelines manually is useful. Running them automatically is where the real leverage comes from.

GridWork HQ includes 12 pre-configured cron jobs. For prospecting, you can set up a weekly run that keeps your pipeline full without you thinking about it.

How Cron Works

The pipeline server uses cron-config.json to define scheduled tasks. Each cron job has a schedule (standard cron syntax), a command to execute, and optional parameters.

From the dashboard, the Cron page shows all 12 configured jobs with their schedules, last run time, and status. You can enable, disable, or modify jobs directly.

Example: Weekly Prospect Sweep

Set up a cron job to prospect a specific industry in your area every Monday at 8am:

{
  "id": "weekly-prospect",
  "schedule": "0 8 * * 1",
  "command": "prospect dentists in Atlanta",
  "enabled": true
}

Every Monday morning, fresh leads appear in your Notion database and the output is archived in the knowledge vault. You review the results over coffee and decide who to reach out to.

Stacking Pipelines

The real power comes from chaining scheduled pipelines. A typical weekly automation might look like:

• Monday 8am: Prospect pipeline runs for your target industry
• Tuesday 8am: Review Monday's leads, trigger outreach pipeline on the best ones
• Friday 4pm: Friday update pipeline generates weekly summaries for active clients
• Monthly: Scope audit pipeline checks all active projects for scope creep

The Tuesday step still requires your judgment (picking which leads to contact), but everything else runs unattended.

What Comes After Your First Pipeline

Once you have the prospect pipeline running and producing useful leads, the natural next steps are:

Run the audit pipeline on hot leads. Type audit [business URL] and get a detailed technical assessment you can send as a free resource. It covers mobile responsiveness, performance, SEO, accessibility, and security. This is your foot in the door.

Try the outreach pipeline. outreach [business] pulls context from the prospect and audit results to draft a personalized email. It references specific issues found on their site, not generic "I can help your business" copy.

Set up the content pipeline for a client. Once you have a paying client, content [client] generates a month of blog posts, social media content, and email newsletters using the client brief in your knowledge vault.

Explore the full set of 17 pipelines. Proposal generation, site build planning, brand identity systems, SEO analysis, monthly reporting, and operational maintenance. Each one is a structured workflow that reads your knowledge vault and produces deliverable output.

Forty-Five Minutes Well Spent

Here is the time breakdown:

• Prerequisites check: 5 minutes
• Install and setup wizard: 10 minutes
• Start both servers: 2 minutes
• First prospect pipeline run: 10 minutes
• Review output and explore dashboard: 8 minutes
• Configure a cron job: 5 minutes
• Run a second pipeline (audit): 5 minutes

Total: 45 minutes from ZIP file to automated lead generation.

The difference between this and chat-based AI is that the pipeline keeps running. Next Monday, fresh leads. Every Friday, client updates. Every month, scope audits. You set it up once and the system works in the background while you focus on the work that actually requires you.

---

Liam builds tools for solo agency operators at GridWork Digital. GridWork HQ is available at gridwork.dev for $199 one-time.