Architecture · AI Architecture for B2B Sales

This is Part 3 of the SignalLeading AI Architecture for B2B Sales series.

In the previous articles, I explained two important ideas:

• AI is like a clever but naive salesperson — it needs a good process, not just good prompts.
  
• SignalLeading is built as a collection of workflows instead of one giant AI application.

In this article, we'll install n8n Community Edition with Docker and build your first simple workflow. You'll have a working environment ready for the more advanced AI workflows in the next articles.

This naturally leads to the next question.

There are many workflow platforms. We chose n8n because it fits the way we think.

1. Visual Design

Every workflow becomes a diagram. Instead of reading hundreds of lines of code, we can immediately see:

• where data starts
• what AI is doing
• which services are involved
• how errors are handled

When we are experimenting with new sales workflows, this visibility is far more valuable than writing everything from scratch.

2. Easy to Test

Good AI systems are built one step at a time. We rarely create an entire workflow in one attempt.

Instead, we test every node. For example:

Receive a company name
↳ Search LinkedIn and other signals in parallel
↳ Filter noise and generate an ICP summary
↳ Review the result
↳ Continue to the next step

Each node can be tested independently. This dramatically reduces debugging time.

3. AI Is Only One of the Components

Many demonstrations treat AI as the centre of the application. Our architecture is different. AI is simply another worker inside the workflow—that is, one of many nodes.

There are many kinds of nodes in a workflow:

• Some nodes call AI.
• Some query databases.
• Some split the flow.
• Some update CRMs.
• Some send emails.

The workflow orchestrates them all.

4. Easy to Integrate

As mentioned, enterprise environments already contain many systems:

Microsoft 365 | Google Workspace | HubSpot | Salesforce | SharePoint | Slack | REST APIs | SQL

n8n already provides connectors for many of these services. That makes expansion much easier as SignalLeading grows. Our goal is not to replace existing corporate systems. Our goal is to orchestrate them.

5. Perfect for Proof of Concept

The Community Edition of n8n runs perfectly on a laptop.

That makes it ideal for:

• validating ideas
• testing AI prompts
• refining workflow logic
• demonstrating concepts to clients

Before investing in production infrastructure, we can first confirm that the business process itself actually works.

This keeps experimentation inexpensive and fast.

---

Installing software directly onto your operating system often leads to configuration problems. Different versions. Different dependencies. Missing libraries.

Docker solves this. Everything runs inside an isolated container on your development laptop. If something breaks, simply remove the container and start again.

First, download Docker Desktop for your operating system. For macOS

1. Open: Double-click the downloaded .dmg file.
2. Drag: Pull the Docker icon into the Applications folder shortcut in that window.
3. Wait: Let the progress bar finish copying the files.
4. Launch: Open your Applications folder and click Docker to initialize it.

You can verify the installation by opening a Terminal and running:

docker --version

If Docker returns its version number, everything is ready.

On macOS, you will notice the Docker whale icon in the menu bar at the top-right of the screen, indicating that Docker Desktop is running.

---

We use the Community Edition, which is the free, self-hosted version of n8n. It is more than enough for learning, prototyping, and building proof-of-concepts on your own laptop.

Unlike Docker, there is no .dmg installer to download. Instead, open a Terminal and execute:

docker volume create n8n_data

docker run -it --rm \
  -p 5678:5678 \
  -v n8n_data:/home/node/.n8n \
  n8nio/n8n

After a few moments, Docker Desktop will show a running n8n container.

Open your browser and visit:

http://localhost:5678

Create your account for n8n.

You now have your own local automation platform.

Stopping n8n

When you have finished working, return to the Terminal window where n8n is running and press:

Ctrl + C

This stops the n8n container.

If you also want to close Docker Desktop completely, click the Docker whale icon in the macOS menu bar and choose Quit Docker Desktop.

Don't worry—your workflows and settings are still safe because they are stored in the Docker volume, n8n_data, not inside the temporary container.

Starting n8n Next Time

The next time you want to continue developing, simply:

1. Start Docker Desktop.
2. Wait until Docker finishes starting (the whale icon becomes stable).
3. Open the Containers dashboard in Docker Desktop, locate the n8n container, and click the Start (▶) button.
4. Once the container is running, open your browser and visit http://localhost:5678.

Your previous n8n instance, credentials, and workflows will be loaded automatically from the n8n_data volume.

---

Don't start with AI. We will discuss how SignalLeading uses AI in the next few articles in this series.

Start playing with something even simpler. Create a workflow containing three nodes:

Manual Trigger
↓
Edit Fields (Set)
↓
Convert to File

1. Node: Manual Trigger

Click Create workflow.

Click the + sign in the middle of the canvas.

On the right-hand panel, search for: Manual Trigger.

The Manual Trigger node starts the workflow when you click Execute Workflow.

2. Node: Edit Fields (Set)

Click the + sign on the right side of the Manual Trigger node.

On the right-hand panel, search for Edit Fields (Set).

Select Edit Fields (Set).

Inside the Edit Fields (Set) node, go to Mode and select JSON.

Add:

{
  "company": "SignalLeading"
}

Click Execute Step.

The result appears on the right:

company = SignalLeading

Two nodes are prepared so far:

3. Node: Convert to File

Click the + sign on the right side of the Edit Fields (Set) node.

On the right-hand panel, search for Convert to File.

Select Convert to File.

In Text Input Field, enter:

{{$json.company}}

You can also drag the company field into this box.

Click Execute Step.

On the Output side, you can see data with the file name file.txt.

Two buttons appear in the output box:

• Click View to see SignalLeading.

  

Or you can choose to:

• Click Download to download the file.

Open the downloaded file.txt.

The word SignalLeading is now saved in the file.

Congratulations.

You have just executed your first n8n workflow.

More importantly, you have seen how a workflow can produce a real result that integrated with your simple workflow.

Although this example only creates a text file, the same workflow could just as easily write to a database, call an API, update a CRM, send an email, or interact with other enterprise systems.

That is the power of workflow orchestration.

Many AI projects begin with prompts.

We begin with workflows.

Workflows reveal:

• where information comes from
• who makes decisions
• what should be automated
• what still requires human judgement

Only after that do we decide where AI belongs.

AI is simply one of the many nodes that can participate in a workflow.

This approach makes the solution easier to test, easier to explain, and easier to evolve.

---

Now that the development environment is ready, we can start building reusable business components.

In the next article, we will build our first real module:

Product & ICP Profiler

Instead of manually entering a company name, we will teach the workflow how to understand a company, identify its products, and describe its Ideal Customer Profile (ICP) in a structured, reusable way.

That profiler becomes the foundation for every workflow that follows in SignalLeading.

#AIArchitecture#B2BSales#n8n#Docker#WorkflowAutomation#AIEngineering#SignalLeading#BuildInPublic