What to Automate First (A Simple Framework)

The most common thing I hear from business owners considering AI automation is: "I know there's stuff we should automate, but I don't know where to start."

That's a reasonable place to be. Every business has dozens of tasks that feel like they should be automatic by now. Invoices. Follow-ups. Reporting. Data entry. Scheduling. The list is long and everything feels important when you're the one doing it all.

The mistake is trying to solve everything at once. The second mistake is picking the wrong thing first, spending money on a project that technically works but doesn't move the needle, and concluding that automation isn't worth it.

The right starting point isn't the most exciting project. It's the one that pays for itself fastest and builds confidence for everything that comes after.

The three-question filter

When I sit down with a business owner for a discovery session, I'm listening for the answer to three questions. You can ask yourself these right now.

1. What costs you the most time every week?

Not what annoys you the most. What actually consumes the most hours. There's often a difference. The thing that annoys you might take 20 minutes but ruin your mood. The thing that costs you time is the four hours of invoice data entry that you barely notice anymore because it's just "part of the job."

Map your week honestly. Where do the hours actually go on tasks that are repetitive, predictable, and don't require your expertise? That's your candidate list.

2. What's the cost if you don't fix it?

Some manual tasks are annoying but harmless. Others are actively costing you money, clients, or opportunities.

Slow lead response costs you jobs. Manual invoice processing delays your cost visibility by weeks. Inconsistent follow-up means customers who would have come back don't. Staff interrupting you with questions that are answered in a document somewhere costs you focused time you never get back.

Put a rough number on it. Hours per week multiplied by the hourly cost of whoever's doing the work. Or revenue lost per month from missed leads or delayed decisions. It doesn't need to be exact. It needs to be honest.

3. Is the data accessible?

This is the practical filter that most people skip. Automation needs data to work with, and that data needs to be in a format a system can read.

If your invoices arrive as PDFs in email, that's accessible. If your leads come through a platform with an API or webhook, that's accessible. If your financial data lives in Xero or MYOB, that's accessible.

If the critical information lives in someone's head, in handwritten notes, or in a system with no way to export data, automation becomes much harder. Not impossible, but it changes the scope and cost significantly.

The best first project scores high on all three: it consumes real time, the cost of not fixing it is tangible, and the data is already digital and accessible.

Common first projects by industry

After dozens of engagements, the same starting points come up repeatedly. Not because I steer people toward them, but because they consistently score highest on the three-question filter.

Hospitality: Invoice processing. High volume, repetitive, error-prone, and the data is already in PDFs. Typical saving: 5 to 10 hours per week.

Trades and services: Lead response. Enquiries arrive after hours, competitors respond first, jobs are lost. The data comes through platforms with webhooks. Typical impact: capturing leads that were previously missed entirely.

Events and entertainment: Cross-platform reporting. Multiple ticketing platforms, ad accounts, and project tools that don't talk to each other. The data is in APIs. Typical saving: 30 to 45 minutes every morning, plus better decisions from cross-platform visibility.

Professional services: Email classification and document search. High volume of incoming communication, staff interrupting each other to find information that exists somewhere in the system. The data is in Google Workspace or similar. Typical saving: reclaimed focus time across the team.

Any business with an owner bottleneck: A persistent AI agent that answers operational questions from staff using actual company documentation. Frees the owner from being the answer to every question, without giving staff access to systems they shouldn't see.

What "start small" actually means

"Start small" doesn't mean "start with something trivial." It means pick one specific problem with a clear outcome, build the solution, measure the result, and use that result to decide what comes next.

A good first project takes two to four weeks to deploy, costs a fraction of the annual time it saves, and gives you concrete data on what automation can do for your specific business. Not a hypothetical case study from someone else's industry. Your numbers. Your workflow. Your result.

That result is what makes the second project an easy decision. And the third. Each one builds on the confidence and infrastructure from the one before.

What not to start with

A few things I actively steer clients away from as first projects:

Anything that requires changing how the whole team works. The first project should slot into existing workflows, not demand new ones. Change management is hard. Save it for later when the team has seen automation work and trusts it.

Anything where the ROI takes months to measure. You want to see the impact in weeks, not quarters. That fast feedback loop is what builds momentum.

Anything where the data is messy or missing. If the first project requires a data cleanup exercise before automation can even start, it's the wrong first project. Pick something where the data is already in decent shape and save the cleanup for a later phase.

"We want AI" without a specific problem. If you can't articulate what changes when the project is done, it's not ready to be a project. That's fine. A discovery conversation will usually surface the real problem within 30 minutes.

The starting point

Pick the task that wastes the most time, costs the most money, and sits on data you already have in a digital format. That's your first project.

If you're not sure which one that is, that's what a discovery session is for. Thirty minutes of honest conversation about your operations will usually surface it clearly.