So far in this programme, you've used AI tools one at a time. You've typed a prompt, received a response, and done something with the output. That's powerful, but it's also a bit like having a very capable colleague who only works when you tap them on the shoulder and give them a specific instruction. What if they could just get on with things?
That's the idea behind AI automation. Instead of using AI tools in isolation, you connect them to the other applications you already use (your email, your calendar, your spreadsheets, your project management tools) so that AI can do useful work in the background, triggered by events rather than by you sitting at a keyboard.
This isn't science fiction, and it isn't just for developers. Automation platforms like Zapier and Make have been connecting apps together for years, using simple "when this happens, do that" logic. What's changed recently is that AI has been woven into these platforms, so the "do that" part can now involve intelligent processing: summarising, categorising, drafting, extracting information, or making decisions based on content rather than just shuffling data from one place to another.
What we mean by automation
At its simplest, automation means setting up a workflow that runs without you having to do anything each time. You define the trigger (something that starts the workflow) and the action (something that happens as a result). A classic example: when someone fills in a contact form on your website, automatically add their details to a spreadsheet and send them a confirmation email. No AI involved; just moving data from A to B.
Where AI enters the picture is in the processing step between trigger and action. Instead of just passing raw data along, an AI step can read an incoming email and summarise it, categorise a support request by urgency, draft a personalised reply based on the content, or extract specific details from an unstructured document. The automation handles the plumbing; the AI handles the thinking.
The tools
Zapier is the most widely used automation platform, connecting to over 8,000 apps with more than 30,000 possible actions. Its free tier gives you 100 tasks per month and lets you create unlimited workflows (called "Zaps"), though each Zap on the free plan is limited to two steps (one trigger and one action). Zapier has also introduced MCP (Model Context Protocol) support, which allows AI chatbots like Claude and ChatGPT to directly trigger actions in your connected apps through conversation. We'll come back to that.
Make (formerly Integromat) takes a more visual approach, with a drag-and-drop canvas where you build workflows as flowcharts. Its free tier offers 1,000 operations per month with two active scenarios, though operations are consumed by every step in a workflow, so they can be used up faster than you might expect. Make is particularly good for more complex workflows with branching logic, but has a steeper learning curve than Zapier.
For the activity in this Thing, we'll focus on Zapier because it's the most accessible for beginners, but if you're curious about Make, it's well worth exploring too.
MCP and practical examples
You may have noticed throughout this programme that AI chatbots are mostly self-contained. You type something, they respond, and that's the end of it. The output stays in the chat window unless you manually copy it somewhere.
MCP (Model Context Protocol) is a standard that changes this. Developed by Anthropic (the company behind Claude) and now an open standard managed by the Linux Foundation, MCP allows AI tools to connect directly to external applications. Think of it as giving your AI chatbot hands as well as a brain. Instead of just telling you what to do, it can actually do it.
In practice, this means you can set up a connection between Claude or ChatGPT and your Zapier account, and then ask the AI to take actions in your other apps using natural language. "Add an event to my Google Calendar for next Tuesday at 2pm" or "Send an email to the team about Friday's meeting" or "Create a new row in my project tracker spreadsheet." The AI understands what you're asking and uses Zapier's connections to make it happen.
This is still a relatively new capability, and it's worth noting that it requires you to grant your AI tool access to your other applications, something to think carefully about before setting up. But it does change what AI chatbots are capable of.
Some practical examples
To make this more concrete, here are the kinds of workflows that combine automation and AI:
None of these require any coding. They're assembled by choosing apps, defining triggers, and configuring AI processing steps through a visual interface.
A word about what can go wrong
Automation is powerful precisely because it runs without your supervision. That's also what makes it risky. A poorly configured workflow can send embarrassing emails to the wrong people, overwrite important data, or rack up costs if it triggers more often than you expected. AI adds another layer of unpredictability. The summarisation might miss something important, the categorisation might be wrong, or the drafted reply might say something you'd never actually say.
The general advice is: start small, test thoroughly, and keep humans in the loop for anything that matters. Use automation for internal processes and low-stakes tasks first. Watch what it does for a while before trusting it with anything public-facing or irreversible. And always check what permissions you're granting when connecting apps, particularly on free tiers, where your data usage terms may differ from paid plans.
Resources to explore
The most widely used automation platform, connecting over 8,000 apps. Free tier includes 100 tasks per month with unlimited two-step Zaps. New accounts get a 14-day Pro trial with multi-step workflows.
A visual automation builder (formerly Integromat) with a drag-and-drop flowchart interface. Free tier includes 1,000 operations per month with two active scenarios. Good for more complex workflows with branching logic.
Zapier's Model Context Protocol integration, which lets AI chatbots like Claude and ChatGPT take actions in your connected apps through natural language conversation.
Activity: build your first AI-powered automation
In this activity, you're going to create a simple automation using Zapier's free tier that includes an AI processing step. You'll connect two apps with an intelligent step in between.
-
Sign up and explore. If you don't already have a Zapier account, create one at zapier.com. When you sign up, you'll automatically get a 14-day trial of the Pro plan, which gives you access to multi-step Zaps and additional features. After the trial, you'll revert to the free plan (100 tasks per month, two-step Zaps only).
-
Create your AI-powered Zap. Build an automation that uses AI to process content between a trigger and an action.
- Test and refine. Once your Zap is configured, use Zapier's built-in testing to run it with your sample data. Check the output: did the AI summarise the content usefully? Is the email or spreadsheet update formatted correctly? If the AI output isn't quite what you want, adjust your prompt. This is the same iterative prompting skill you've practised throughout this programme, and it works just the same in an automation context.
-
Explore Zapier MCP (optional). If you use Claude (on a paid plan that supports integrations) or ChatGPT, you can try connecting Zapier's MCP server to give your AI chatbot the ability to take actions directly.
-
Reflect on the experience. Think about what just happened and what it means.
Your output for this Thing should include:
- A screenshot of your completed Zap (showing the trigger, any AI step, and the action)
- A brief description of what your automation does and what prompted you to choose it
- If you tried MCP, a screenshot or description of the experience
- A short reflection covering what worked, what you'd automate next, and what risks you'd want to manage
Why this matters
For most of this programme, you've interacted with AI as a conversational partner. You ask, it answers. That's useful, but it's also limited. It depends on you being there, asking the right questions, at the right time.
Automation changes the model. Instead of AI as a tool you pick up and put down, it becomes something more like a process that runs alongside your other work. The AI doesn't replace you; it handles the routine processing that would otherwise eat into your time, and flags the things that need your attention.
This is also where the concept of "agentic AI" starts to become real. Throughout 2025 and into 2026, the major AI companies have been building towards systems that don't just respond to prompts but take multi-step actions on your behalf. MCP is one piece of that puzzle, a standard way for AI tools to interact with the rest of your digital life. Browser-based AI agents (like Claude's Chrome extension or OpenAI's Operator) are another, and we'll touch on those in Thing 22.
The practical skill you've developed here (thinking about your work as a series of triggers, processes, and actions, and identifying where AI can add value in that chain) is likely to become increasingly important. Even if the specific tools change (and they will), the underlying pattern of connecting AI to workflows is only going to become more common.
It's also worth being clear-eyed about the limitations. Free tiers are useful for learning and for light personal use, but they're not designed for heavy or business-critical workflows. Automation platforms charge based on usage, and costs can escalate quickly as you add more workflows or process more data. And as with any tool that accesses your accounts and data, security and privacy should be front of mind, particularly when granting AI tools permission to take actions on your behalf.
Claim your Open Badge
Submit your screenshot of your completed automation workflow, a description of what it does, your optional MCP exploration, and your reflection on the experience via cred.scot.
Submit your automation screenshot and reflection to claim this badge via cred.scot.
Claim now