Think in jobs, not tools
Start from what the business must do each week, then pick the lightest tool for each job. A stack of five jobs is enough to begin:
| Job | Example options | Check before you commit |
|---|---|---|
| Publish | Blogger, a static site, a hosted website builder | Can you export your content? Can you use your own domain? |
| Collect contacts | A form that writes to Google Sheets, an email service's signup form | Where is the data stored, and who can access it? |
| Automate | Google Apps Script, n8n, Make, Zapier | Task or run limits on the free plan; self-hosting licence terms |
| Report | Looker Studio with Google Sheets or GA4 | Data freshness and credentials (see our Sheets guide) |
| Draft with AI | A local open-weight language model | Hardware, quality, human review |
What "free" really means
- Free tiers change. Prices, limits and features are decided by each vendor and can change, so read the current pricing page on the day you decide and write the date down.
- Free usually has a ceiling: a number of tasks, contacts, runs or storage. Estimate your monthly volume first.
- Self-hosted is not automatically free. It costs hardware, electricity, your time and security upkeep. Check the licence too: some "source available" tools restrict commercial use.
- You pay in lock-in. Prefer tools that let you export data in plain formats such as CSV.
A working example from this site
This site's contact form is a small stack in itself: a form on a Blogger page sends the message to a Google Apps Script web app, which appends a row to a Google Sheet and emails a notification. We covered the build and the mistakes in our troubleshooting log. It uses only Google's free services, and its limit is volume: Apps Script has daily quotas, so a high-traffic form needs a different design. Check Google's current quota page for numbers.
How to choose, in four questions
- Does it do the job today without custom code?
- Can I export my data?
- What happens when I outgrow the free plan, and what does the next plan cost?
- Who can see the data, and does that match my privacy policy?
Security basics
- Use a password manager and two-step verification on every account.
- Give each tool the minimum permissions it needs.
- Keep a list of tools, owners and renewal dates.
- Back up your sheets and automations monthly.
Drafting with a local AI model on your own laptop
In the site owner's own workflow, AI-assisted drafting of summaries and first drafts of customer replies is done with a local AI model running on the owner's own laptop instead of a cloud chat service. This describes the owner's setup, and it is separate from how this article was produced (see below).
- What it is: an open-weight language model that you download once and run through a local runner program. Smaller models need less memory and run on ordinary laptops; larger ones are slower and need more memory.
- Why local: prompts, customer details and drafts stay on hardware you control, there are no per-token fees, and nothing changes underneath you unless you update it. Local does not mean automatically secure, so protect the machine and limit network access to it.
- Trade-offs: speed depends on your hardware (a long draft can take minutes on a laptop), and quality varies by task. Smaller models are more likely to invent details, so give them the facts and nothing else.
- Human review: every draft is edited by a person before it is sent or published. A model can state wrong facts with confidence..
Starter prompt (adapt it):
Summarise the customer message below in two sentences, then draft a polite reply of under 100 words that answers only what was asked. Do not promise refunds, dates or prices. Message: [PASTE MESSAGE]
In an automation tool such as n8n, a node can send this prompt to the local model on your own machine and store the reply, for example as a draft row in a Google Sheet, for a person to review before anything is sent.
What we actually run
The stack described here is close to what we use. This blog is on Blogger, performance is read in Search Console, and our automation workflows were built in the free self-hosted n8n with a small AI model running locally. We ran all nine test workflows on one CPU-only laptop on October 26 and 27, 2026. Each run took roughly one to two minutes, including about 15 seconds for n8n to start. We make no claim about what any of it earns; the point is that the tooling itself did not need a paid plan to test.
How this article was produced
This article was researched and drafted with AI assistance, working from the primary sources listed below, then published by the site owner, who is responsible for it. If you find an error, tell us. Our standards are described in the editorial policy.
What we have and have not verified
The standards and thresholds quoted come from the primary sources linked below. The workflows, checklists and recommendations are our own judgement for small sites, and we are not quoting results or traffic figures because they depend entirely on your situation. We have not benchmarked any of the tools named, and prices are deliberately not quoted.
Sources
- Google Workspace (check current quotas for Apps Script)
- n8n (check the current licence terms)
- Google Search Central: Creating helpful, reliable, people-first content
Found an error or something out of date? Tell us and we will review and correct it.
About the author: Junaid Sabir writes about practical AI and automation for people without a technical background. This article was drafted with AI assistance, then checked and edited by the author; the "What we have and have not verified" notes say which parts were tested.