Start here · Lesson 1.1
AI Workflow Automation for Solo Founders: What Works
If you run a business alone, you have probably lost days to comparing automation tools. Zapier, Make, n8n, AI agents, each with a loud fan base and a different idea of what automation means.
This guide does not rank them with invented scores. It explains the two kinds of tool, what each is good at, and how to choose in an afternoon instead of a month.
Rule based tools run the same steps every time. AI agents handle jobs where the steps change. Most solo founders need a bit of each, starting with whichever matches their most annoying task.
Two kinds of automation
Rule based tools such as Zapier, Make and n8n connect apps with fixed steps. Zapier's own help pages describe workflows that connect your apps to automate repetitive tasks without coding, and n8n describes itself as a workflow automation tool that also has AI features. For example: when a form is filled in, add a row to a sheet, then send an email. They tend to be predictable and good when the job never changes shape. Note that the lines between the two kinds are blurring, since some of these tools now add AI steps.
AI agents such as Claude Code take a goal in plain words and work out the steps. Anthropic's docs describe Claude Code as reading files, editing them and running commands, and it can connect to other tools through MCP, an open standard for linking AI to tools and data. That suits a messy folder, a page to edit, or an inbox where every message is different. They are flexible, but you must check their work.
Which one fits which job
| Job | Better fit | Why |
|---|---|---|
| New lead lands in a form, goes to a sheet | Rule based | Same steps every time |
| Send a receipt when a payment clears | Rule based | Fixed trigger, fixed action |
| Decide which emails need a reply | AI agent | Every email is different |
| Edit a landing page and check it | AI agent | Needs to read and change real files |
| Weekly report from messy notes | AI agent | Needs judgement about what matters |
| Move data between two apps nightly | Rule based | Predictable and repeatable |
The test is simple. If you can write the steps on a card and they never change, use a rule based tool. If you would need to say "it depends", use an agent.
The cost that gets ignored: your time to maintain it
Every automation you build is something you must keep working. Rule based flows break when an app changes its fields. Agents drift when your instructions are vague. Either way, fewer automations that you trust beat many you do not.
So before you build anything, ask three questions. How often does this happen? How much does it annoy me? What is the harm if it gets it wrong? Our guide on what to automate first turns those into a scoring sheet.
A one afternoon way to choose
- List the tasksWrite down every repeating task from last week. Aim for ten.
- Mark the shapeTick each one as fixed steps or needs judgement.
- Pick one of eachChoose your most annoying fixed task and your most annoying judgement task.
- Try the matching toolUse a rule based tool for the first and an AI agent for the second. Give each one week.
- Keep what workedDrop the one you did not trust. Do not buy a bigger plan yet.
Where AI agents earn their place
The biggest difference is that an agent works on things that are not neatly structured. A folder of documents, a half finished website, an inbox. In the Halftime lesson on setting up Claude Code the examples are cleaning up a messy desktop, building a landing page, drafting inbox replies and building dashboards from your numbers. None of those fit neatly into a trigger and an action.
An agent also lets you change the job by talking. Changing a rule based flow means opening a builder and editing blocks. Changing an agent's job means saying what you want differently.
Where rule based tools still win
Do not replace everything with an agent. A payment receipt, a calendar reminder or a data sync should be boring and exact. A rule based tool will do those the same way a thousand times. An agent might phrase something differently on run number 400.
Guardrails that apply to both
- Start with tasks you can undo. A draft is safer than a sent email.
- Keep a human check on anything that touches customers or money until it has been right many times.
- Write down what each automation does and who owns it, even if the owner is you.
- Keep secrets such as passwords out of anything an agent can read.
What to do this week
Pick the one task that irritates you most and apply the test above. Do not build a system. Build one small thing, run it for a week, and look at what it saved you. If it helped, do the next one.
Questions to ask any tool before you commit
Whatever you are considering, ask what it does when something goes wrong. Does it stop and tell you, fail silently, or retry forever? Silent failure is the most expensive kind, because you find out when a customer complains.
Ask who can see your data and where it is stored, and read the privacy page rather than the sales page. Ask what leaving looks like: can you export your flows or your files, or are you locked in? Ask what the pricing does when you grow, because many tools charge by the number of runs and a busy month can surprise you. Check the current numbers on each vendor's own site, since plans change.
A simple example of using both
Say you sell a small online course. A rule based tool can watch for a new purchase and add the buyer to your email list and a spreadsheet. That is fixed and should stay boring. An AI agent can then read the last week of support emails and draft a short list of the questions people keep asking, which you turn into a new lesson page. The first is exact and repeated. The second needs reading and judgement. Neither replaces the other.
When to stop adding tools
If you spend more than an hour a week fixing automations, you have too many. Remove the one you trust least and see whether anything is missed. Often nothing is. A smaller set that works beats a clever set that needs you.
Questions people ask
Is n8n or Zapier better than an AI agent?
They solve different problems. Rule based tools are best for fixed steps between apps. Agents are best when the job changes each time or works on files.
Can I use both together?
Yes, and many founders do. A rule based tool can trigger the routine parts while an agent handles the parts that need judgement.
What should a solo founder automate first?
A task that repeats often, annoys you, and is easy to undo if it goes wrong.
Will an AI agent make mistakes?
Sometimes. That is why you start with reversible work and check the output until you trust it.
Next step
If you want to see an agent doing real work on a computer, Lesson 1.1 sets one up with you from scratch.