How to automate your job: the practical guide

Which tasks automate and which ones stay with you

Not every part of your job is automatable. The dividing line is clear once you see it.

The dividing line is consistent. Automatable tasks repeat, follow a predictable structure, and produce an output you could recognize as right or wrong within 30 seconds. Non-automatable tasks require judgment that changes depending on the situation, the person, or something you learned five minutes ago.

The leverage in automating your job comes from the first column. Most solo operators spend two to three hours a day on tasks in that column. Getting those hours back is the point.

The three questions that find your automatable tasks

Before you touch any tool, run every task in your week through three questions.

Does it repeat? At least weekly, or every time a specific thing happens: a form submission, a new booking, a payment received. Tasks that only come up once or twice a year are rarely worth automating, even if they feel painful when they arrive.

Is there a pattern? The same kind of input leads to the same kind of output, every time. A lead submitting the enquiry form always gets the same first email. A new week always produces the same Monday report. The template stays the same even if the names and numbers change.

Can you describe what "right" looks like? If someone showed you the output, you'd know in about 30 seconds whether it was good. That means you can tell the AI when it is wrong. That feedback loop is the whole mechanism that makes automations work. If you cannot describe the right result, you cannot correct the AI when it drifts, which means the automation will not hold up.

Yes to all three: worth automating. If any one is no, move on. The highest-leverage thing you can do in the first week is get good at spotting the tasks that pass all three tests, because that is where most of your time is sitting untouched.

The five tasks small business owners automate first

These are the most common starting points because the patterns are tightest and the volume is highest.

1. Follow-up email sequences

When a new lead or customer takes an action (submits a form, books a call, makes a purchase), a sequence of emails goes out automatically. The first is immediate. A second goes two days later. A third on day seven. Each email is templated with the right name, the right context, and the right next step for where they are in the relationship.

This is the most common first automation because it runs all the time, used to require real attention, and the cost of skipping it (leads going cold, new customers with no follow-through) shows up immediately in your numbers. Most people who set this up for the first time realize they were personally following up with about 30% of the people who should have heard from them.

2. Weekly business reports

Every Monday morning, a summary of last week arrives in your inbox. New leads. Revenue. Which channels sent traffic. Which emails got replies. Which deals moved. The numbers come from your own database. The writing happens automatically. You read it in three minutes and know the state of the business without opening a single dashboard.

Before this automation exists, that report either gets produced manually over an hour of copying numbers from different places, or it does not get produced at all. Most solo operators fall into the second category. The report that never gets made is the one where you find out something is broken two weeks after you could have caught it.

3. Booking confirmations and reminders

When someone books a session, they get a confirmation with the date, the time, and what to prepare. Twenty-four hours before, they get a reminder. Both fire automatically when the calendar entry is created. No copy-pasting, no checking whether you sent one, no one slipping through without a reminder because you had a busy day.

4. Proposal first drafts

When a lead reaches a certain point in your pipeline (a call completed, a budget confirmed, a service tier selected), an AI agent pulls their company name, their stated problem, and the right pricing tier, then generates a proposal draft. You spend five minutes reviewing and personalizing it before sending. The draft that used to take 45 minutes now takes five. The difference is not that you are cutting corners. It is that the first draft is no longer blank.

5. Lead routing and CRM updates

When a new contact lands in your CRM, an agent checks the source, scores the lead against your ideal customer criteria, assigns a tag, and sets a follow-up reminder. You open the CRM on Monday and every new lead is already organized, already scored, already ready to work. The sorting you used to do at the start of every week disappears into the background where it belongs.

How to hand a task to AI: the three-step loop

The method for setting up any automation is the same, regardless of what the task is.

Step 1: Write the task as steps. Before you involve any tool, write out what you actually do when you perform this task: what triggers it, what you do first, what you do next, what the output looks like when it is right. Three to ten sentences is usually enough. This is the brief the AI works from. Skipping this step is what produces automations that look right in the first test and break in production.

Step 2: Hand it to the AI and get a first draft. Give the AI your written description and the relevant context: a sample email, a template you have used before, a note about what good looks like. Ask it to produce the output. You are looking for a first attempt, not a finished result. Expecting a finished result on round one is the wrong frame.

Step 3: Review and adjust. Read the output against your brief. If something is off, say what it is and ask for a fix. "The tone is too formal" or "this is missing the client's company name in the subject line" is enough. Most tasks reach a working state after two or three rounds. Once the output is consistently good across a week of examples, the automation is ready to run unattended.

The loop is simple by design. The discipline is in step one: writing the task out before you open a tool. People who skip the description step tend to spend five rounds of feedback getting back to what they would have said in the first brief.

The automations running inside SoloStack right now

These are real automations inside SoloStack's own in-house stack, all running for about $13 a month in infrastructure costs.

Workshop enquiry follow-up: When someone submits the enquiry form, three emails go out automatically over seven days. Confirmation on day one. The most common objection answered on day three. A simple "still thinking it over?" on day seven. The sequence runs on Resend (about $10/mo for email delivery) and a Netlify function. No Klaviyo. No Mailchimp. No monthly platform fee tied to subscriber count.

Monday lead summary: Every Monday at 8am, an agent queries the CRM database, counts leads from the previous week, groups them by source (referral, search, social, direct), and emails a three-paragraph summary. The whole thing generates and sends itself. Ten minutes to set up. It has run every week since without anyone touching it.

Booking confirmations and 24-hour reminders: When someone books a session, a webhook fires and two emails are queued: immediate confirmation and a reminder the morning before. Both pull the right name, date, and prep notes from the booking record. No one has ever received a manual confirmation email from SoloStack.

Proposal drafts: When a qualified lead reaches the proposal stage in the CRM, an agent generates a draft using their company name, stated problem, and the product tier they discussed. The team reviews it and edits for five minutes before sending. The draft that used to take 45 minutes now takes 30 seconds to generate and five minutes to polish.

None of these would have been realistic to build three years ago. All of them are now a few hours of setup, each. The infrastructure is commodity. The skill is knowing which task to describe first.

What to tackle after the first five

Once the operational automations are running, the next wave is usually the content layer.

Taking one blog post and repurposing it into a LinkedIn carousel. Turning notes from a customer call into a case study draft. Generating a weekly content calendar from what performed well last week. These automations take a bit longer to set up because the patterns are less rigid than email sequences. But the payoff is higher because they remove work that would otherwise eat your whole afternoon and get skipped when the week gets busy.

After the content layer, most people move to research and reporting: competitor monitoring, keyword tracking, lead enrichment, and scoring. By that point the mindset is second nature. The question stops being "can this be automated?" and starts being "how long will it take me to describe this one clearly?"

The businesses that run lean are not the ones that found some magic tool. They are the ones that got disciplined about separating the repeating work from the judgment work, and handed over the first column as fast as they could.

SoloStack can help you map your working week to the automations worth building first and set them up with you, live, in the workshop. We have run the same audit on our own operations, and we walk you through it on yours.