The Marketers & Managers Guide to Building AI Agents with n8n (No Coding Required)

Key Takeaways

• UK businesses are not adopting AI as fast as you might expect. The DSIT UK AI Adoption Survey (2025) found that only 16% are currently using AI, and OpenAI’s own research suggests that among those who do, 80% are still using basic chat rather than automated workflows
• AI agents are not chatbots – they range from semi-autonomous systems with a human in the loop to fully automated workflows, but the key difference is that they trigger, decide, and act across your business tools rather than just generating text in a window
• We have developed a model called the “AI Adoption Ladder”, which allows you to identify where you are in the use of automation and AI. We recommend that you start at level 2-3 (AI-enhanced automations with a human in the loop) rather than trying to jump straight to fully autonomous agents. This will build confidence with your supervised workflows before removing the human checkpoints
• n8n is our preferred platform for building AI automations, though alternatives like Zapier and Make are also available. You can start using n8n for free, upgrade to paid cloud-hosted plans for convenience, or get a techie/developer to install the open-source self-hosted version – 230,000+ users and a $2.5 billion valuation show this is a serious platform
• You do not need to be a developer to start – the Joke Machine example, described below, proves anyone can build a working AI agent in under 10 minutes. The cloud version makes connecting Gmail and other platforms easy, though you will need to learn about credentials and API keys to link other platforms and tools. For more sophisticated automations, code nodes may be needed, but tools like Claude can write those for you
• At Anicca, we run AI agents through our secret-agents.ai platform (triggered by forms that anyone can use) and internally for meeting note transcriptions, weekly blog creation, and automated client reports – this is not experimental, it is how we operate every day
• As you progress up the AI Adoption Ladder, you are likely to use a newer and more interactive type of agents, such as Claude Cowork, Claude Code or Gemini Antigravity (with others being launched on a weekly basis). Building these automations is easier than automated workflows like n8n – as you can describe what you want in plain English (or even by voice) and it writes the code for you. It is also far more powerful than n8n because it can handle day-to-day tasks like managing your files, searching the web, and posting content on your behalf

---

About This Guide: From ChatGPT to AI Agents and n8n

If you have ever copied text from ChatGPT and pasted it into an email, a spreadsheet, or a social media scheduler, you already understand the problem. You are the glue between AI and your business tools. The AI thinks, but you still have to do all the doing – copying, pasting, reformatting, uploading. AI agents change that. They connect the thinking directly to the doing, so the work happens without you sitting in the middle.This guide walks you through what AI agents actually are, how they differ from the chatbots you already use, and how to build your first one using n8n – a visual automation platform that does not require any coding. You do not need to be a developer. I am a marketer, and Anicca’s other director (Darren Wynn) and I have built over 30 AI workflows at our agency without writing a single line of code.I recently gave a workshop on this topic at the Agency Hackers “The Robots Are Coming” conference at the British Library. The audience was agency owners, marketers, and managers – most of whom were regular users of ChatGPT and were building CustomGPTs, but had not yet built anything automated. This article expands on that workshop into the guide I wish I had been given when I started.

Most Businesses Are Stuck at ChatGPT

Let me start with a statistic that might surprise you. OpenAI’s own enterprise data from 2025 shows that 80% of business AI usage is still basic chat – people asking questions in a text box and copying the answer into a document. Only 20% of usage flows through custom GPTs or Projects.The UK picture is even more stark. The Department for Science, Innovation and Technology (DSIT) found that only 16% of UK businesses are currently using AI, with a further 5% planning to adopt. Among sole traders, 42% have no plans to adopt AI at all.And yet, among the businesses that have adopted AI, 75% report improved productivity. The gap between “using ChatGPT” and “using AI agents” is where the real gains live. This guide is about crossing that gap.

The AI Adoption Ladder

I use a model I call the AI Adoption Ladder to help people understand where they are and where they need to get to. There are five rungs:

Rung 1: Basic Automation but without AI. This is the sort of automation you may have already used in a marketing role – within HubSpot, Salesforce, Klaviyo, or another CRM. You witness it every day as a consumer too. You buy something online and immediately receive a thank you email with your order confirmation, dispatch details, and a request to share on social media. If a WhatsApp message follows shortly after, that is an API integration at work – where tools like Zapier connect the e-commerce platform to the retailer’s WhatsApp account, so it can message you automatically.Rung 2: Basic ChatGPT or Gemini. This is where the majority sit. They are asking questions, rewriting text, maybe generating the occasional social media caption. It is useful, but it is manual – every interaction requires a human to start it, guide it, and do something with the result. Unfortunately the output can often be poor because of the user’s lack of prompting skills.Rung 3: Custom GPTs and Claude Projects. This is where you start getting serious. You save your prompts, upload context documents like brand guidelines or product data, and create reusable AI setups with specific instructions. Claude Projects take this further with persistent memory – you can return to previous conversations and the AI remembers what you discussed. You are building a brain that makes the AI smarter about your business over time. But it is still one person working with one AI in one window.Rung 4: Automation tools like n8n. This is the step change. You are connecting AI to your real business systems – your CRM, your email, your spreadsheets, your social platforms. The AI does not just think; it acts. These automations can be semi-autonomous, with a human triggering them or reviewing the output, or fully autonomous, running on a schedule or firing when someone submits a form. They are reliable and predictable, but they are not interactive – they do not learn or evolve as you use them.Rung 5: AI agents and Claude Code. This is where AI stops following predefined workflows and starts creating new things. Tools like Claude Code can read your entire project, write and edit files, search the web, and build tools that did not exist before – all from a plain English description or even a voice instruction. Google’s Antigravity and Manus offer similar capabilities. You can also build reusable skills – packages of instructions that teach the AI how to perform specific tasks repeatedly, like generating a client report or writing a blog post in your brand voice.At Anicca, we have built over 80 skills and use them daily to produce audit reports, interactive portals, and marketing collateral that would take a team days to complete manually. Unlike rung 4, these agents are interactive. They learn your preferences, remember your brand guidelines, and get better the more you work with them.Most businesses think they are “using AI” because they sit on rung 2. The real productivity gains start at rung 4. This guide is about getting you there.

What Is an AI Agent (and What It Is Not)

An AI agent is not a chatbot. A chatbot generates text inside a conversation window. You type, it responds, you type again. An AI agent is fundamentally different: it uses a large language model (an LLM – the technology behind ChatGPT and Claude) to understand inputs, make decisions, use tools, and complete multi-step tasks autonomously.I explain this using what I call the Christmas Cracker model, because every AI agent has three parts that pull apart like a cracker:

• The Trigger – what starts the process. This could be a form submission, a new email arriving, a scheduled time, or a webhook from another application.
• The Agent – the AI brain that decides what to do. This is the LLM (Claude, GPT, Gemini) processing the information and making decisions.
• The Output – the action the agent takes. Send an email, update a spreadsheet, post to social media, create a document, notify your team on Slack.

That said, not every agent runs without supervision. Many of the most useful agents are semi-autonomous – they do the heavy lifting but pause at key points for a human to check and approve before continuing. The fully automated version is the end goal, not necessarily the starting point. The important thing is that the agent handles the structured, repetitive work so you are only involved where your judgement genuinely matters.For marketers and managers, the shift in thinking is this: stop asking “what can AI write for me?” and start asking “what would I hire a junior employee to do?” – then build that as an agent.

The Agent Spectrum: Six Types from Simple to Autonomous

Not all AI agents are the same. I use a spectrum of six types to help people understand the range of what is possible – and where to start:

Type 1: Simple Automation. If X happens, then do Y. No AI involved – this is traditional automation like Zapier triggers. A new row in a spreadsheet sends an email. Simple, reliable, but not intelligent.Type 2: AI-Enhanced Automation. The same trigger-action model, but with an AI step in the middle. For example: a new lead arrives, AI classifies it by industry and urgency, then routes it to the correct sales rep. The AI adds intelligence to an otherwise mechanical process.Type 3: AI Workflow. Multiple steps where AI makes decisions at each stage. A blog brief arrives, AI researches the topic, generates a draft, checks it against your brand guidelines, and schedules it for publication. Each step involves AI judgement.Type 4: Supervised Agent. The agent runs autonomously for most of the process but asks for human approval at key decision points. It drafts client emails but sends them to you for a final check before dispatching. This is where most agencies should aim to start.Type 5: Autonomous Agent. Runs end-to-end without human intervention. Monitors your Google Ads performance, identifies underperforming campaigns, adjusts budgets, and sends you a summary report – all without you touching anything.Type 6: Personal Assistant. Always on, context-aware, learns from your patterns over time. Knows your schedule, your preferences, your clients, and your communication style. This is where the technology is heading.The key message is: you do not need to jump to type 6. Start at type 2 or 3 with n8n. Most agencies will get massive value from types 3 and 4, and you can build those today.

Automation You Already Know: HubSpot, GoHighLevel, and the Limits

If you have used HubSpot, Salesforce, Klaviyo, GoHighLevel, or any CRM with a workflow builder, you have probably built these kinds of automations yourself. Email sequences that fire when a form is submitted, lead scoring rules that move contacts between lists, nurture campaigns that drip content over time. Most of this is built on “if this happens, then do that” logic – classic conditional switching. It works, and it is the foundation that most marketing automation is built on.These platforms have evolved significantly. Many now include their own AI features, built-in integrations, and the ability to connect to third-party tools via webhooks and Zapier. For some businesses, this will be all the automation they ever need, and there is nothing wrong with that.The limitation comes when you want the automation itself to think. Traditional workflows follow the same path every time regardless of context. An AI-powered workflow can read the incoming data, make a judgement call, and adapt its response accordingly. The AI node can analyse a lead’s behaviour, check their company size, read their latest LinkedIn post, and write a genuinely personalised follow-up – all within the same workflow. That is the difference between a sequence and an agent, and it is why platforms like n8n exist.

Why n8n? The Tool We Chose

There are several automation platforms available – Zapier, Make (formerly Integromat), and n8n being the three most popular. We chose n8n for our agency work, and here is why:

It is visual. You build workflows by dragging and dropping nodes on a canvas and connecting them with lines. You can see exactly how data flows from one step to the next. There is no code to write unless you want to.It has flexible pricing. You can start with n8n’s free tier to learn the platform. When you are ready to scale, cloud-hosted plans give you the convenience of managed infrastructure. And if you have technical support available, the open-source self-hosted version means you can run n8n on your own server – keeping sensitive client data on your own infrastructure.

It treats AI as a first-class citizen. n8n has a dedicated AI Agent node with built-in memory, tool calling, and guardrails. Zapier treats AI as “just another integration.” n8n was designed with AI workflows at its core.It has 400+ integrations. Google Sheets, Gmail, Slack, HubSpot, Salesforce, Airtable, Notion, WordPress – if you use it, n8n probably connects to it. And if it does not, the HTTP Request node lets you connect to any API.The numbers speak for themselves: n8n has 230,000+ active users, raised $180 million in its Series C at a $2.5 billion valuation (October 2025), and has over 4,000 community-contributed workflow templates.And the results are real. Vodafone deployed 33 n8n workflows and saved 5,000 person-days, avoiding £2.2 million in costs. Their ongoing savings run at approximately £300,000 per month. As the team at Huel put it: “n8n was the big unlock. Tools like ChatGPT and Claude are great, but n8n is the thing that allows you to integrate AI into your work and your processes in a safe and controlled way.”But you do not need to be a Vodafone or a Huel to see the benefits. At Anicca, we use n8n in two ways. Externally, we built secret-agents.ai – a platform where clients and contacts can trigger AI agents by filling in a simple form, no technical knowledge required. Internally, we run over 30 workflows for things like email triage, meeting note transcriptions, blog creation, and automated client reporting. Most businesses can start with something just as small – a single workflow that saves 30 minutes a day adds up to over 120 hours a year.How does it compare? Zapier is easier for absolute beginners and has 8,000+ integrations, but it is significantly more expensive at scale and has limited AI capabilities. Make has a polished interface and is great for simpler automations. n8n wins on complex workflows, AI agent capabilities, self-hosting, and cost. The typical progression is: start with Make, graduate to n8n as your needs grow.

n8n Core Concepts: The Four Building Blocks

Before building anything, you need to understand four concepts. These are the building blocks of every n8n workflow:

Workflows. A workflow is a series of connected steps that run automatically. Think of it as a recipe – first do this, then do that, then do the other thing. Each workflow has a starting point (a trigger) and one or more actions.Nodes. Each step in a workflow is called a node. A node is a single action – send an email, read a spreadsheet, call an AI model, post to Slack. You drag nodes onto the canvas and connect them to build your workflow.Triggers. Every workflow needs something to start it. A trigger could be a schedule (run every morning at 9am), a webhook (another application sends a signal), a new email arriving, a form submission, or a manual button press. The trigger is always the first node.Data Passing. When a node runs, it produces data. That data gets passed to the next node in the workflow, which can use it, transform it, or pass it along. Data flows through n8n in a format called JSON – but you do not need to understand JSON to use it. The visual interface shows you exactly what data is available at each step.

Build Your First Agent: The Joke Machine

When you sign up to n8n Cloud and log in, this is your workflow dashboard. From here you can create new workflows, manage credentials, and see your execution history.

When you create a new workflow, you are presented with two options: add your first step manually, or use the AI Workflow Builder to describe what you want in plain English and let it build the workflow for you.

At the conference, I built the joke machine below, live in under 10 minutes, to prove that anyone can do it. The Joke Machine is a 3-node workflow that takes someone’s name and a topic, generates a personalised joke using AI, and emails it to them. It is simple, but it demonstrates every principle of an AI agent.Node 1: Form Trigger. n8n has a built-in form node that creates a simple web form. Ours asked for three things: the person’s name, email address (so we know where to send the joke) and a topic they wanted a joke about. When someone submits the form, the workflow starts.Node 2: AI Agent (Anthropic Chat Model). This node takes the name and topic from the form and sends them to Claude with a prompt: “Generate a short, corny personalised joke for [name] about [topic], in the style of Billy Connolly.” Claude generates the joke and passes it to the next node.Ann, a dog walks into a bar. The barman says “We don’t serve dogs in here.” The dog says “That’s fine, I’ll have a pint of lager. I don’t eat dogs either.”Node 3: Gmail. The final node takes the generated joke and sends it by email to the person who submitted the form. Subject line, body text, and recipient are all pulled automatically from the previous nodes.

Three nodes. No code. A working AI agent that takes input, thinks, and acts.

A note on getting started: One of the advantages of n8n’s cloud version is that connecting to platforms like Gmail and Google Sheets is straightforward – you simply log in with your Google account. However, as you progress to more integrations, you will need to learn about credentials and API keys – the authentication tokens that let n8n talk to other services on your behalf. This is a learning curve, but not a steep one, and n8n’s documentation walks you through each connection step by step.And when you need more sophistication? Some advanced workflows require Code nodes for custom logic – data transformations, conditional formatting, or complex routing. But here is the key insight: you do not need to write that code yourself. Tools like Claude can generate the code for you when you describe what you need. In fact, n8n itself now has an AI Workflow Builder that can create entire workflows from a natural language description. You describe what you want, and it builds the workflow for you – you just need to connect your own credentials.The Joke Machine is intentionally simple, but the pattern is identical for serious business workflows. Replace the form with a CRM webhook, replace the joke prompt with lead qualification criteria, replace the email with a Slack notification – and you have a lead scoring agent. The building blocks are the same.

Real Business Workflows You Could Create Now

The Joke Machine proves the concept. Here is what it looks like when you apply the same principles to real marketing work:

Lead Qualification and Routing. A new enquiry comes in through a website form. The AI agent analyses the company name, checks their website, scores them against our ideal client profile, and routes the lead to the correct team member with a Slack notification and a pre-written email draft. What used to take 15 minutes per lead now happens in seconds.Content Repurposing Pipeline. A new blog post is published. The workflow triggers automatically, reads the full article, and generates: a LinkedIn post, an email newsletter summary, and a set of carousel bullet points – all tailored to each platform’s format and character limits.Client Performance Reporting. Every Monday morning, a scheduled workflow pulls data from Google Analytics and Google Ads, feeds it to Claude for analysis, formats the insights into a branded report template, and emails it to the client. No human touches it unless the AI flags something unusual.Email Triage and Drafting. You can run a Gmail bot that runs every 30 minutes. It reads new emails, classifies them by urgency and topic using Claude, drafts appropriate responses, and sends Teams notifications to the right person. It runs 24 hours a day, 7 days a week.

The Christmas Cracker Model: Thinking About AI Employees

Once you have built a few workflows, the bigger picture starts to emerge. I use the Christmas Cracker model to explain how AI agents become the equivalent of AI employees in your business:

Inputs – everything that feeds into your business: voice notes, emails, client briefs, data feeds, calendar events, social media mentions, form submissions, CRM updates.The Content Engine – your n8n workflows and AI models working together. This is the brain of the operation. It takes raw inputs, processes them through multiple AI steps, applies your business rules, and generates outputs.Outputs – everything your business produces: blog posts, social media content, client reports, email responses, proposals, internal summaries, performance dashboards.The shift in thinking is profound. Instead of asking “what tool should I use?” you ask “what employee do I need?” A social media manager? Build a content agent. A lead qualifier? Build a scoring agent. A reporting analyst? Build a data agent. Each one runs continuously, never takes a day off, and costs a fraction of a salary.This does not replace people. It replaces the repetitive, structured work that stops your people from doing the strategic and creative work they were hired to do.

The 10 Nodes That Cover 80% of Use Cases

n8n has hundreds of nodes, but you do not need to learn them all. These 10 will cover the vast majority of what marketers and managers need to build:

With just these 10 nodes, you can build lead qualification systems, content pipelines, reporting automations, email triage bots, and much more.

Nodes that help in the creation of content and marketing assets

• Document handling Many of our workflows involve creating or processing documents – meeting transcriptions, proposals, first drafts, client reports. The Extract Content node lets you pull text from uploaded files such as PDFs and Word documents so the AI can read and work with the contents. On the output side, binary file nodes allow you to create downloadable documents from your workflow’s results. Between the two, you can build workflows that take a raw document in, process it through AI, and produce a finished file at the other end.

• Image Creation. Recent models from ChatGPT and Gemini’s Nano Banana have significantly improved the ability to generate and edit images within an automated workflow. In practice, we find it more reliable to start with a branded template and ask the AI to edit it rather than generating images from scratch, as prompting for visual output within an n8n workflow can be unpredictable.This does introduce some technical steps. Image editing and image generation use different API endpoints (the URL that connects your workflow to the AI service). For example, when creating a carousel, an ad, or a blog featured image, we use ChatGPT’s image edit endpoint and pass it variables from the workflow – the blog title, the category, the author name – so it knows what to change on the template. The output comes back as a base64 string (a long block of encoded text rather than a visible image), so you need an additional node to convert that into a PNG file and save it somewhere you can actually use it.

• HTML Display. Most automations produce their output as raw JSON data or plain Markdown text, neither of which is easy to read or present to a client. The HTML node lets you take that raw output and format it into a clean, styled page – whether that is a generated report, a blog draft for review, or a preview of the image you just created. We use this node in almost every workflow because it turns something that looks like code into something a human can actually read and act on.

When You Need Technical Support

You can build a lot with n8n without any technical help, but there are areas where a developer or someone more technical adds real value:

• Setting up the self-hosted open-source version on your own server
• Running workflows as background processes that start automatically when the server boots
• Configuring direct integrations with your email system or CRM at the API level rather than through pre-built nodes
• Locking down access controls, SSL certificates, and security compliance when handling sensitive client data
• Setting up error monitoring and alerts so you know immediately when a workflow fails rather than discovering it days later
• Building user authentication, billing integration, and scalable infrastructure if you want to turn a workflow into a product or SaaS offering

None of this should stop you starting on your own – but it is worth knowing where the boundary sits between what you can do and where professional support pays for itself.

Beyond n8n: When Visual Automation Is Not Enough

n8n is powerful, but it has a ceiling. Visual drag-and-drop workflows are perfect for defined, repeatable processes – lead routing, content scheduling, email triage, reporting. But there are tasks that need more flexibility.When you need AI that does not just connect existing tools but creates new ones – building a bespoke client report from scratch, generating a 40-page audit with custom charts, or creating an interactive portal – that is where conversational AI agents come in.This is not about choosing one or the other. At Anicca, we use both:

• n8n for defined, repeatable workflows (email triage, lead scoring, content distribution, reporting)
• Conversational AI agents for open-ended, creative work (building audit reports, generating proposals, creating client tools)

No-code (n8n) gives you speed. You can prototype a workflow in an hour, test it in a day, and have it running in production by the end of the week. Anyone on your team can build and maintain it. Changes take minutes.Conversational AI agents give you depth. You interact with them the same way you would chat with a colleague – describing what you want in plain English. Behind the scenes, they create the output using code, which gives you far more functionality than any drag-and-drop tool. They can read and work with your existing files, so they understand your business context from the start. You can build entire applications, bespoke client deliverables, and complex data transformations without writing a line of code yourself.A fast-moving landscape. This space is evolving month by month. Anthropic’s Claude Code was one of the first tools to let non-developers build sophisticated outputs through conversation. They followed it with Claude Cowork, designed specifically for non-technical office workers and connecting directly to Google Drive, Gmail, and enterprise tools. Google launched Antigravity, a full AI-powered development environment. Microsoft has integrated Claude into its Copilot Cowork tier.The way these agents share and extend their capabilities is also changing rapidly. OpenClaw emerged as an open-source marketplace for AI agent skills, connecting to platforms like WhatsApp, Slack, Discord, and Teams. It gained massive popularity – over 200,000 stars on GitHub – but its open nature has raised significant security concerns about unvetted skills and data handling. In response, Anthropic launched Claude Code Channels, offering similar functionality through Telegram and Discord but with the security and oversight of a commercial platform. Perplexity and Nvidia have launched their own competing marketplaces. The pattern is clear: every major AI company is racing to build the ecosystem around their agents.The investment for you is in learning how to brief these tools effectively, not in learning to program. And the principles you learn with one tool transfer directly to the next.

Understanding the Claude Ecosystem: Seven Levels of Capability

For those ready to go beyond n8n, or who want to understand where the technology is heading, here is how I explain the Claude ecosystem. It has seven levels, each building on the last:1. Claude Chat. The starting point. Conversational AI – you ask questions, it answers. Similar to ChatGPT but with different strengths (particularly in long-form content, reasoning, and following complex instructions).2. Claude Projects. You can create saved workspaces with custom instructions and uploaded documents. This is like giving Claude a permanent briefing document about your business, your brand voice, or your industry.3. Claude Artifacts. Claude can generate documents, code, visualisations, and interactive content directly in the conversation. Not just text – actual usable outputs.4. Claude MCP (Model Context Protocol). This is where it gets interesting. MCP lets you connect Claude to external tools and services – browsers, calendars, databases, APIs. Claude can interact with your local files, not just generate text.5. Claude Code. A conversational AI agent that can read your entire codebase, write and edit files, run commands, and make changes across multiple files. It is like having an AI developer working alongside you.6. Claude Skills. Reusable packages of instructions, scripts, and resources that Claude Code can learn. At Anicca, we have built over 80 skills – from generating SEO audits, to building competitor intelligence decks. Each skill is a capability the AI permanently knows how to perform.7. Claude Channels and OpenClaw. Claude Channels launched this week (March 2026) as Anthropic’s platform for connecting Claude Code to messaging apps like Telegram and Discord, so you can trigger agents and receive results directly in the tools your team already uses.It is Anthropic’s answer to OpenClaw, the open-source AI agent that went viral in early 2026 by letting you message an AI assistant through WhatsApp, Slack, Teams, and more – it runs on your own machine and can read files, browse the web, send emails, and automate tasks across your applications.OpenClaw remains the most widely adopted option with over 200,000 GitHub stars, but its open nature and the broad permissions it requires have raised significant security concerns. Claude Channels offers a more controlled alternative with Anthropic’s safety and security standards built in. Expect this space to change rapidly as every major AI company is building their own version.

Getting Started: Your Next Three Steps

You do not need to go to level 7 of the Claude capability ecosystem to get value. Many businesses will find enormous benefit at levels 1-4. But knowing the full landscape helps you plan where to invest your time and training.If this guide has convinced you to move beyond basic ChatGPT usage, here are the three things I would recommend doing this week:1. Try n8n. Sign up for the free cloud tier at n8n.io and build the Joke Machine. It will take you less than 30 minutes, and you will understand every principle in this article by doing it.2. Pick one workflow to automate. Look at your week and identify one task that is repetitive, structured, and involves moving information between systems. Lead notifications, content distribution, or weekly reporting are good starting points. Build it in n8n.3. Join the Thursday AI Club. We run fortnightly sessions where marketers and agency owners learn to build AI agents hands-on. Visit anicca.co.uk/thursday-ai-club to find out more.For further reading, these guides go deeper into the Claude ecosystem:

• From Claude Projects to Claude Code: A Beginner’s Guide
• 5 Claude Code Skills Every Business Should Build First
• The Marketer’s Guide to OpenClaw 
• Claude Just Got Marketplace, Dispatch and Code Channels in 7 Days – What It Means for Your Business

You can also download the full presentation slides from the Robots Are Coming conference at anicca.co.uk/resources-webinars

Frequently Asked Questions

Do I need coding skills to use n8n?

No. n8n is a visual drag-and-drop platform. You build workflows by connecting nodes on a canvas, not by writing code. There is an optional Code node for advanced users, but it is not required for the vast majority of workflows.

Is n8n free?

n8n has a free self-hosted option where you run it on your own server (as little as £5 per month for hosting). The cloud-hosted version starts at $20 per month. Compared to Zapier, which charges $69+ per month for equivalent functionality, n8n is significantly more cost-effective at scale.

What is the difference between n8n and Zapier?

Zapier is easier for absolute beginners and has more pre-built integrations (8,000+). n8n has fewer integrations (400+) but treats AI as a core feature rather than just another connection. n8n also offers self-hosting for data privacy and is substantially cheaper at scale. For agencies handling client data, the self-hosting option is often the deciding factor.

What is an LLM and why does it matter?

LLM stands for Large Language Model. It is the technology that powers ChatGPT, Claude, and Gemini. These models have been trained on vast amounts of text and can understand, generate, and reason about human language. In n8n, you use LLMs as the “brain” of your AI agents – the node that reads inputs, makes decisions, and generates outputs.

Can I use n8n with client data safely?

Yes. n8n’s self-hosting option means you can run everything on your own infrastructure. Client data never passes through a third-party server. This is one of the primary reasons agencies choose n8n over cloud-only alternatives. You control where the data lives and who has access to it.

What is Claude Code and how is it different from n8n?

n8n is a visual automation platform for connecting existing tools and building repeatable workflows. Claude Code is an AI agent that can read your codebase, write files, run commands, and build entirely new tools and systems. Think of n8n as connecting the plumbing between your apps, and Claude Code as the architect that designs and builds new rooms. Most agencies start with n8n and graduate to Claude Code as their ambitions grow.

Glossary

• AI Agent – an autonomous system that uses an LLM to perceive inputs, make decisions, and take actions without constant human guidance. Different from a chatbot, which only responds when prompted
• AI Adoption Ladder – a five-rung framework for understanding where a business sits in its AI journey, from not using AI at all (rung 1) to deploying autonomous agents (rung 5)
• Anthropic – the AI company that makes Claude. Founded by former OpenAI researchers, headquartered in San Francisco
• API (Application Programming Interface) – a way for two software applications to talk to each other. When n8n connects to Google Sheets, it uses Google’s API. You do not need to understand the technical details – n8n handles the connection for you
• Automation – making a process run without manual intervention. Email sequences in HubSpot are automation. n8n workflows are automation. The difference is that n8n can add AI intelligence to the process
• Christmas Cracker Model – Ann Stanley’s framework for explaining AI agents: every agent has a Trigger (what starts it), an Agent (the AI brain), and an Output (the action it takes)
• Claude – Anthropic’s AI assistant, used in the n8n AI Agent node. Known for strong performance in long-form content, complex reasoning, and following detailed instructions
• Claude Code – a command-line AI agent from Anthropic that can read your entire codebase, write and edit files, run commands, and build new tools. The next level beyond visual automation
• Context Window – the amount of information an LLM can hold in memory during a single interaction. A larger context window means the AI can work with longer documents and more complex tasks
• JSON (JavaScript Object Notation) – the data format used to pass information between nodes in n8n. You do not need to write JSON – n8n displays it visually – but knowing it exists helps when debugging workflows
• LLM (Large Language Model) – the technology behind ChatGPT, Claude, and Gemini. A type of AI trained on enormous amounts of text that can understand, generate, and reason about human language
• MCP (Model Context Protocol) – a way to connect Claude to external tools and services such as browsers, calendars, and databases. It allows the AI to interact with real-world systems, not just generate text
• n8n – an open-source workflow automation platform (pronounced “n-eight-n”). Lets you build automated workflows by visually connecting nodes on a canvas. Has 230,000+ users and a $2.5 billion valuation
• Node – a single step in an n8n workflow. Each node performs one action: send an email, read a spreadsheet, call an AI model, or post a message. You connect nodes to build workflows
• No-Code – building software or automations without writing programming code, using visual interfaces instead. n8n is a no-code platform
• OpenClaw – an open-source marketplace for sharing and discovering Claude Code skills. Like an app store for AI agent capabilities
• Prompt – the instructions you give to an AI model. In n8n, you write a prompt inside the AI Agent node to tell Claude or GPT what to do with the incoming data
• Self-Hosting – running software on your own server rather than using someone else’s cloud. n8n can be self-hosted, meaning your data stays on your infrastructure
• Skills (Claude Code) – reusable packages of instructions, scripts, and resources that Claude Code can learn and perform repeatedly. Each skill is a permanent capability
• Trigger – the first node in an n8n workflow that starts the process. Can be a schedule, a webhook, a form submission, a new email, or a manual button press
• Webhook – a way for one application to send real-time data to another when an event occurs. In n8n, a webhook trigger starts your workflow when an external app sends it a signal
• Workflow – a series of connected nodes in n8n that automate a process from start to finish. The equivalent of a recipe: trigger, steps, output

This article is based on Ann Stanley’s workshop at the Agency Hackers “The Robots Are Coming” conference, British Library, London, 19 March 2026. Ann is the CEO of Anicca Digital, a Leicester-based digital marketing agency specialising in AI-powered marketing automation.