Automate Lead Generation and Weekly Marketing Reports with n8n: 2 Real Workflows

Short answer

The pattern behind both workflows

Look past the tools and both workflows do the same five things a marketing assistant would do by hand:

1. Start on time: every day at 8:00, or every Monday at 9:00 (Schedule Trigger).^[1]
2. Collect: call an API and receive JSON (HTTP Request).^[2]
3. Clean and decide: keep only useful items, combine sources, cap the volume (Code, Merge, Filter, Limit).
4. Write: let a language model draft the email or the analysis (AI Agent or Basic LLM Chain).
5. Act and remember: send the email and record it so tomorrow's run doesn't repeat it (Gmail, Google Sheets).

Each step passes a list of items to the next, and every node runs once per item automatically.^[3][4] If that sounds abstract, our first article explains how n8n items and expressions work. Here we stay at the business level: no code, just what each node is for.

Workflow 1: e-commerce lead generation, every morning

What it does. The workflow has two branches that run one hour apart.^[5] At 8:00 it finds shops in Tunisia that have a website listed in OpenStreetMap. It visits each website, extracts a public contact email, removes addresses already in the list and appends the new leads to a Google Sheet. At 9:00 it reads the sheet and keeps only the shops not yet contacted, takes ten of them, asks an AI model to write a personalised email for each one, sends it and marks the row as sent. If anything fails, the error-alert workflow from part 2 of this series sends an email.

The lead nodes, explained in general

Three design choices worth copying

• Split discovery from outreach. Two branches an hour apart mean a slow scrape never delays emails, and you can review or clean the sheet in between.
• The sheet is the memory. The “sent” flag makes each run aware of the previous ones, so nobody gets the same email twice.
• Cap the volume. A Limit node is the simplest safety valve against sending hundreds of AI-written emails because of one bad filter.

Workflow 2: a weekly marketing report written by AI

What it does. Every Monday at 9:00, the workflow asks Google Analytics 4 three questions about the last seven days: which city sent the most visitors, how the site performed overall (sessions, users, bounce rate), and which pages were read most.^[5] It sends this to a language model, which returns a short strategy report and a draft outreach email. The report is emailed to the site owner. The workflow then looks up businesses with a website in the top city and sends them the outreach email. It connects “where our visitors come from” to “who we should talk to”.

The report nodes, explained in general

• HTTP Request to an analytics API. GA4's Data API has a runReport method. You ask for dimensions (city, page) and metrics (sessions, views) over a date range, and it returns rows.^[16] Any analytics, ads or e-commerce platform with an API works the same way.
• Code nodes as translators. Each API answers in its own format. A short Code step after each call turns the answer into simple named fields, such as city, sessions and bounce rate, that the next nodes and the AI can read.^[7]
• Basic LLM Chain vs AI Agent. The Basic LLM Chain sends one prompt and returns one answer;^[17] the AI Agent can decide to call tools and loop.^[12] A weekly report needs no tools, so the simpler chain is cheaper and more predictable. Use the agent when the model must look things up.
• Asking the AI for structured output. The prompt asks the model to answer in a fixed shape (report, email subject, email body), and a Code node checks it before anything is sent. Never pass raw AI output straight to customers.
• Gmail to the owner first. The human gets the analysis before any outreach goes out. That's the review step.

In Part B, the HTTP Request nodes call OpenStreetMap's Nominatim service to turn the city name into map coordinates, then Overpass to list businesses in that area.^[18][6] These are free community services with strict usage rules, which we cover below.

Apply it to your business

Keep the skeleton (schedule → collect → clean → write → act and remember) and change the sources and the message:

When you have many clients or properties to go through, the same nodes simply run once per item. When you must group results, such as one report per client with totals, the Split Out, Aggregate and Summarize nodes take over.

Stay legal: data, email and API rules

Automating outreach makes it easy to break rules at scale, so build them into the workflow from day one. This is general information, not legal advice.

• Personal data needs a legal basis. Under the EU GDPR, processing personal data requires a lawful basis, and people can object to direct marketing at any time, after which you must stop.^[19] Add an opt-out line and a “do not contact” column that your Filter node respects.
• Marketing email rules. The EU ePrivacy Directive generally requires prior consent for marketing email to individuals, with a narrow exception for existing customers.^[20] Generic company addresses carry less risk than named personal ones, but check the rules in your target country.
• Tunisia has its own law. Organic Law 2004-63 governs personal data processing in Tunisia,^[21] so the same care applies to Tunisian leads.
• Respect the data sources. OpenStreetMap data is under the Open Database License and requires attribution.^[22] The public Nominatim service allows at most one request per second, requires an identifying User-Agent, and asks you to cache results.^[18]
• Keep a human in the loop. Review AI-written emails before volume grows. A model can invent facts about a business it never visited.

What the research says about the gains

The writing steps in these workflows are where language models help most. A study published in Science estimated that about 80% of the US workforce could have at least 10% of their tasks affected by large language models, and about 19% at least half of them.^[23] In a study of 5,172 customer-support agents published in The Quarterly Journal of Economics in 2025, access to an AI assistant raised the number of issues resolved per hour by 15% on average, and less experienced workers improved in both speed and quality.^[24]

The automation steps add a different gain: consistency. The report arrives every Monday even in a busy week, and the follow-up list never forgets anyone. A 2025 field experiment across 66 firms points the same way: workers who used an AI tool spent about two fewer hours a week on email and worked less outside regular hours.^[25]

Next in the series: n8n data pipelines with Split Out, Aggregate, loops and AI agents, based on a sales report and a healthcare data pipeline.