Build an AI News Workflow in n8n with OpenAI (Beginner Guide)

If you think you need to learn Python or work with APIs to build an AI assistant, there’s good news—you don’t. With n8n, you can create your first AI-powered workflow in just 20–30 minutes using ready-made workflow nodes.

In this guide, we’ll build an AI assistant that finds the latest artificial intelligence news, selects the most important stories, summarizes them, and emails you a daily digest. By the end of this tutorial, you’ll understand how to build an AI workflow in n8n and use the same approach to create your own AI automations.

If you’re new to AI assistants, we also recommend reading What Is an AI Agent? AI Agent vs AI Assistant vs AI Automation. It provides helpful background on how these systems work.

What This AI Workflow Does

In this tutorial, an AI workflow refers to the automation you build in n8n, while the AI assistant is the language model that performs the tasks inside that workflow.
Our assistant performs just one task, but simple workflows like this are the best way to learn.

Once it’s running, it will:

• search for the latest AI news;
• filter out less relevant articles;
• generate a concise summary of each story;
• email you a ready-to-read daily news digest.

Don’t try to build a complex AI assistant that handles dozens of tasks right away. It’s much easier to master one focused workflow first, then gradually expand it as your skills improve.

What You’ll Need Before You Start

Before creating your assistant, make sure you have:

• an account at n8n;
• an OpenAI or Anthropic account for connecting a language model;
• Google Chrome, Microsoft Edge, or Mozilla Firefox.

Everything in this tutorial is built using a no-code AI workflow in n8n.

The free n8n plan is more than enough to complete this tutorial and build your first AI assistant.

Step 1. Create a New Workflow

After logging into n8n, click Create Workflow. A blank canvas will open where you’ll build your automation.

For now, don’t add any nodes. Take a moment to explore the interface. The large area in the center is your workflow canvas, and the + button is used to add new nodes.

What you should have at this point: an empty Workflow ready for configuration.

Step 2. Add the AI Agent Node to Your Workflow

Click the + button and search for AI Agent.

Select the node with that name and add it to your workflow.

This node becomes the brain of your AI workflow and powers your AI assistant. All of the key configuration steps in this guide will take place inside it.

If you can’t find the node, make sure you’re using the latest version of n8n.

Step 3. Connect a Language Model

Now it’s time to connect a language model. This is the component that will analyze information, search the web, and generate responses.

The language model powers the entire AI workflow by analyzing information and generating responses.

Open the AI Agent node you just added by double-clicking it. Scroll to the bottom until you find the Chat Model or Model field (the exact label may vary depending on your version of n8n).

After selecting a language model, a window will appear with a button labeled Create new credential or Set Up Credential. Click it.

You’ll need an API key for this step. You can generate one from your OpenAI or Anthropic account. If you don’t have one yet, don’t worry—it only takes a few minutes. We explain the entire process in detail in our separate guide: How to Get an OpenAI or Anthropic API Key.

Once your credentials have been created, select the model you want to use. Any modern model is perfectly suitable for your first project. There’s no need to choose the most expensive model just because it sounds more powerful.

When everything is configured correctly, the Chat Model field will no longer display authorization warnings.

What you should have at this point: your selected language model is connected to the AI Agent node, and there are no credential or authorization errors.

Step 4. Give Your Assistant Clear Instructions

Your AI assistant needs clear instructions describing exactly what it should do. These instructions are provided through a System Message, which the language model reads before processing every request.

While still inside the AI Agent node, scroll to the Options section. Click Add Option and select System Message. A large text field will appear.

Paste the following prompt into the System Message field:

You are an AI news analyst.

Find original artificial intelligence news stories published within the last 24 hours.

Give priority to official sources from the following organizations:

• OpenAI
• Anthropic
• Google DeepMind
• Microsoft
• Meta
• NVIDIA
• Hugging Face

If the same event is reported by multiple websites, always use the original source.

Do not include sponsored content, promotional material, reposts, or rewritten articles.

Whenever possible, prioritize official company blogs, press releases, and other primary sources. Avoid articles that simply quote or summarize the original announcement.

For each news story, provide:

• Title
• A brief explanation of the news (up to 80 words)
• A short explanation of why this news matters for users, businesses, or the development of AI. Do not make assumptions that are not supported by the original source.
• A link to the original source

Format every news item using the following structure:

1. Title

Summary:
...

Why it matters:
...

Source:
...

Leave one blank line between each news story.

Do not invent news, dates, quotes, facts, or links.

If you cannot find five verified news stories published during the last 24 hours, return only the stories that you can confirm through web search.

Reply only in clear, natural English.

If the original source is written in another language, understand its meaning first and then write a natural English summary instead of translating it word for word.

Do not copy the style of company press releases.

Rewrite each story in your own words using a neutral, informative tone.

Avoid marketing language, exaggerations, or promotional wording.

Do not begin your response with phrases such as "Sure", "Okay", "Certainly", or "Here are the results".

Start immediately with the list of news stories.

Do not ask follow-up questions or suggest additional actions after the last news item.

End the response after the final news story.

Format the entire response in Markdown.

Use headings, lists, and blank lines to improve readability.

Before submitting your response, make sure the text is grammatically correct and free of spelling or stylistic errors.

After pasting the prompt, you don’t need to save anything manually. n8n saves changes automatically.

What you should have at this point: your prompt is visible inside the System Message field, and the AI Agent node is fully configured for this stage.

Don’t modify the prompt before your first test run. First, make sure your assistant works correctly with these instructions. Once everything is working, you can experiment by refining the prompt and adding new requirements.

Step 5. Enable Web Search for Your AI Workflow

To work with current information, your AI assistant needs access to OpenAI’s built-in web search tool.
This allows your AI workflow to retrieve up-to-date information instead of relying only on the model’s training data.
Open the OpenAI Chat Model node.

Make sure Use Responses API is enabled.

Under Built-in Tools, click Add Built-in Tool and select Web Search.

After enabling Web Search, several additional search settings will become available.

For your first AI assistant, you can safely leave the default settings unchanged. Later, you can restrict searches to specific websites, regions, or search scopes if needed.

Step 6. Test Your AI Workflow

Click the Open Chat button.

A chat window will appear at the bottom of the screen. Since your workflow is currently triggered by chat, simply type any message—for example, Start—and send it.

This message is only used to trigger the workflow. Your AI assistant will immediately begin searching for the latest AI news and processing the results.

The generated output will appear in the panel next to the chat window.

The first run usually takes a little longer than subsequent ones. That’s perfectly normal.

If everything has been configured correctly, you’ll receive a result similar to this:

1. OpenAI releases a new feature...

Summary...

Why it matters...

Source: https://...

Don’t worry if your results look slightly different. News changes constantly, so every execution is likely to return different stories.

What to Do If Your Assistant Doesn’t Work

In most cases, problems are caused by configuration issues rather than n8n itself. Before looking for complex solutions, check the following:

Three Simple Improvements That Make Your Assistant Much More Useful

Once your first AI workflow is running successfully, you may be tempted to build something much more complex. Resist that urge.

Instead, improve your assistant one feature at a time. Test every change before moving on to the next one.

1. Limit the Sources

If you only want news from trusted websites, simply add another instruction to your prompt.

Use only official company websites or reputable technology publications.

This significantly reduces the chances of your assistant using questionable sources.

2. Change the Output Format

If you plan to publish the summaries on your blog, Telegram, Slack, or anywhere else, ask the assistant to generate the content in the format you need.

For example:

Format the entire response in Markdown.

This saves a lot of editing time later.

3. Add an Importance Score

Another useful improvement is asking your assistant to rate each story.

After every news story, assign an importance score from 1 to 10 and briefly explain your rating.

When dozens of stories are published every day, this makes it much easier to decide which ones deserve your attention first.

Five Tips That Will Save You Hours

• • Don’t create a brand-new assistant for every small change. Improve the one you’ve already built.
  • Modify only one part of your prompt at a time. If you change several things simultaneously, it becomes difficult to understand what affected the output.
  • Save your best prompts in a separate document. Over time, you’ll build your own prompt library that you can reuse in future projects.
  • Test your workflow after every change. Don’t move on until the current version works reliably.
  • Don’t try to build a universal AI assistant. A focused assistant designed for one specific task almost always performs better.

How to Use This AI Assistant in Real Work

Once your AI workflow is working, don’t leave it as a simple learning exercise. With just a few prompt changes—or by adding another tool—you can turn it into something genuinely useful in your daily workflow.

Here are a few practical ideas.

The great thing is that you don’t need to build a brand-new AI assistant every time. In most cases, changing the prompt is enough to completely transform what it does.

Ready-to-Use Prompts

One of the fastest ways to understand what AI assistants can do is to experiment with different prompts.

AI Assistant for Finding Jobs

You are a career assistant.

Every day, find newly published AI Automation Engineer and n8n Developer job openings.

Only include jobs posted within the last 24 hours.

For each position, provide:

• Company
• Country
• Short description
• Salary (if available)
• Link to the original job listing.

AI Assistant for Finding Blog Ideas

Find the most interesting artificial intelligence news published during the last 24 hours.

For every news story, suggest three original blog post ideas.

The titles should be SEO-friendly and written for Google Search.

AI Assistant for Competitor Research

Analyze the latest articles published on the specified website.

Identify the main topics.

Suggest five article ideas that could outperform the competitor.

Once you learn how to change only the prompt, you’ll quickly realize that a single AI assistant can perform dozens of completely different tasks.

Checklist Before Your First Run

Before moving on to more advanced automations, make sure you’ve completed every step.

• • • ✅ Created a new Workflow.
    • ✅ Added the AI Agent node.
    • ✅ Connected a language model.
    • ✅ Enabled Web Search.
    • ✅ Added a detailed System Message.
    • ✅ Completed a successful test run.
    • ✅ Received the expected results without errors.

If everything on this checklist is complete, you’re ready to take your workflow one step further by scheduling it to run automatically and sending the results directly to your inbox.

Let’s improve our AI workflow.

Step 7. Replace Manual Execution with an Automatic Schedule

So far, you’ve been running the workflow manually. That’s perfect for testing, but in real-world use it’s much more convenient to let n8n run it automatically.

At the beginning of your workflow, locate the first node. In our example, it’s called When chat message received.

Select it and either press the Delete key or remove it using the node menu.

Next, click the + button to open the list of available nodes.

Search for Schedule Trigger and add it to your workflow.

Double-click the node to open its settings.

Under Trigger Interval, choose Days.

New scheduling options will appear.

Set the Hour field to 9 and the Minute field to 0.

This tells n8n to run your workflow automatically every day at 9:00 AM.

After that, connect the output of the Schedule Trigger node to the input of the AI Agent node.

If the connection wasn’t created automatically, click and hold the small circle on the right side of the Schedule Trigger node, drag the line to the circle on the left side of the AI Agent node, and release the mouse button.

Your AI workflow is now fully automated.

Every morning at 9:00 AM, n8n will launch your assistant automatically. It will search for the latest AI news, summarize it, and pass the results to the next step in the workflow.

What you should have at this point: the first node in your workflow is now Schedule Trigger, configured to run every day at 9:00 AM, and connected to the AI Agent node.

7.1 If You See the “No Prompt Provided” Error

After replacing the When Chat Message Received node with a Schedule Trigger, your AI assistant no longer receives a user message automatically.

Because of this, the first execution may fail with the error “No prompt provided”.

To fix it, open the AI Agent node.

At the top of the settings, locate the Source for Prompt (User Message) option.

By default, it’s set to Connected Chat Trigger Node, which means the assistant is still waiting for a message from the old chat trigger.

Click this field and select Define below.

Once you do that, the Prompt (User Message) field becomes editable.

Enter the prompt you want your assistant to execute every time the workflow runs automatically.

For example:

Prepare today's AI news digest.

From now on, every scheduled execution will automatically send this request to your AI assistant, which will then follow the instructions defined in the System Message.

What you should have at this point: Source for Prompt (User Message) is set to Define below, and the Prompt (User Message) field contains the request that should run automatically.

Step 8. Automatically Email Your AI News Digest

The final step is to have your assistant email the news summary automatically.

Once this is configured, you won’t need to open n8n every morning. The workflow will run on schedule and deliver the results directly to your inbox.

Click the + button to the right of the AI Agent node, or click the small + icon on its output.

Search for Gmail and choose the Send a message action.

If this is your first time using Gmail in n8n, click Sign in with Google, sign in to your Google account, and grant the requested permissions.

Once authorization is complete, your credentials will be selected automatically.

Leave Resource set to Message.

Leave Operation set to Send.

In the To field, enter the email address where you’d like to receive the daily digest.

For the Subject field, you can use something like:

Daily AI News Digest

Now it’s time to insert the output generated by your AI assistant into the email.

Click Add Expression inside the Message field.

The Expression Editor will open.

From the left panel, select the AI Agent node, then drag the output field into the Message box.

n8n will automatically generate the correct expression.

Save your changes and click Execute Workflow to test the complete automation.

If everything is configured correctly, your AI assistant will search for the latest AI news, generate a summary, and send it to your email address within a few seconds.

What you should have at this point: after a successful test, you’ll receive an email containing your AI-generated news digest. Once the workflow is activated, a new digest will arrive automatically every day at 9:00 AM.

Step 9. Activate Your Workflow

Your AI workflow is fully configured, but it won’t run automatically until it’s activated.

In the upper-right corner of the n8n editor, click Active (or Activate, depending on your version of n8n). If your workspace uses the newer interface, click Publish.

Once the workflow is activated, n8n will keep it running in the background.

From this point on, every day at 9:00 AM, the workflow will start automatically, search for the latest AI news, generate a summary, and send the results to your email.

What you should have at this point: the workflow is active, and the status indicator confirms that it is ready to run automatically according to the schedule.

FAQ

Can I build AI workflows without coding?

Yes. That’s exactly what no-code platforms like n8n are designed for. Most of the work is done by connecting visual workflow nodes and writing prompts instead of code.

Is the free n8n plan enough?

Absolutely. The free plan is more than sufficient for learning, testing workflows, and building your first AI assistant. Paid plans become useful only when you need larger-scale or production automations.

Which language model should I choose?

Any modern model is suitable for getting started. Learning how to write effective prompts is much more important than choosing the most expensive model.

Why isn’t my assistant working correctly?

Most issues are caused by an overly generic prompt or a missing tool such as Web Search. Before troubleshooting further, verify that your language model is connected and your workflow is configured correctly.

Can I build multiple AI workflows?

Yes. Most people eventually create several AI workflows, each designed for a specific task. One assistant can monitor news, another can summarize documents, while another manages email or customer support. Smaller AI automations are easier to maintain than one large workflow.

What to Try Next

Building your first AI workflow teaches one of the most important lessons about AI automation: powerful workflows are usually built from small, well-designed steps.

Instead of trying to automate everything at once, keep improving your existing workflow one feature at a time.

Once you’re comfortable with n8n, try extending this project by creating assistants that automatically post updates to Telegram, summarize documents, monitor Gmail, organize information in Notion, or generate ideas for future content.

After your first successful automation, you’ll discover that building the next one becomes much easier.