You know you should be doing something with AI in your business.
Every week, another email lands from a vendor. Another LinkedIn post from someone announcing their breakthrough. Another board member asks what your AI strategy is. Another competitor seems to be moving faster than you on something you have not even started.
The pressure is real. The paralysis is also real. The question is not whether AI matters for your business. It does. The question is where, exactly, you start, when there are forty plausible entry points and you can only really pursue one or two well.
A 2024 Shopify merchant survey found that the primary barriers to AI adoption are not technical. They are knowledge gaps. Twenty-nine percent of business owners are unclear on what AI can actually do. Another twenty-nine percent know what it can do but are not sure how to take the first step. This is the most common operational stuck point in business right now, and it is the focus of this post.
The good news is that the right first move is more obvious than it looks. The bad news is that it is almost never the one your vendor will recommend, the one your board is most excited about, or the one the news cycle is fixated on this week. The right first move comes from inside your business, not from outside it.
The wrong way to start
Before getting to the framework, it is worth naming the patterns that keep operations leaders stuck.
Tool-first thinking. You read about a new AI tool. It looks impressive. You sign up. You try to find something in the business to do with it. This is the most common starting pattern and the most common reason AI initiatives die quietly within six months. The tool dictates the use case rather than the other way around.
The most-hyped use case. You start with whatever your industry is currently most excited about. AI agents, chatbots, predictive analytics, the buzzword of the quarter. The problem is that the most-hyped use cases are often the hardest to implement well, the slowest to deliver value, and the most likely to embarrass you publicly if they fail. Hype is a terrible prioritisation framework.
The board-pleasing project. You pick the use case that will look most impressive to your CEO, board, or investors. The trouble is that high-visibility AI projects almost always have unrealistic expectations attached, no operational owner, and no clear path from pilot to production. The combination kills them.
Trying to do everything at once. You map out a comprehensive AI roadmap with twelve initiatives across every function in the business. Three quarters later, none of them have shipped, your team is exhausted, and the roadmap quietly gets archived.
If any of these sound familiar, you are in good company. Most growing businesses have tried at least two of them. The pattern that actually works looks completely different.
The right way to start
The right starting point for AI in your business has three characteristics. It is a real operational problem you already have. It is something where the cost of the problem is measurable. And it is something where success can be verified inside thirty days.
Those three filters knock out almost every initiative that fails. They also point you straight at the right first move.
The single best place to start with AI in almost every growing business is the highest-frequency, highest-cost, highest-judgement-light task on your operations team's plate.
Let me unpack each of those words because they all matter.
Highest-frequency. It happens often. Multiple times a day, ideally. The reason this matters is that AI delivers value through volume. A tool that saves you ten minutes on a task you do once a month is not worth implementing. A tool that saves you ten minutes on a task you do forty times a day is transformative.
Highest-cost. The task is consuming significant senior time or causing significant downstream friction. You can quantify it. "This is costing us approximately twenty hours a week" or "this is delaying every downstream process by an average of two days" or "this is costing us roughly $8,000 a month in salary terms." The quantification matters because it gives you the success criterion. You will know whether the AI is working because the number will change.
Highest-judgement-light. The task does not require deep human judgement, relationship context, or creative thinking. It is mostly mechanical. Pulling data from one place and entering it into another. Drafting a structured response based on a template. Classifying inputs into categories. Summarising standard information. These are the tasks AI is genuinely good at right now. The tasks that require nuanced human judgement are the ones to leave alone for a while longer.
If you can find a task that scores high on all three filters, you have your first move.
The categories worth looking at first
Here are the operational areas where, across hundreds of growing businesses, AI delivers the fastest and most measurable first wins.
Customer and prospect data preparation. Your team is pulling lead information from three different sources, formatting it, deduplicating it, and entering it into your CRM. This is high-frequency, high-cost, and judgement-light. AI handles it well.
Invoice and document processing. Your accounts payable team is keying invoice details into your accounting system, categorising expenses, and matching purchase orders. The bottleneck is data entry, not approval. AI cuts the data entry time dramatically while keeping the approval workflow with humans.
Internal knowledge search. Your team spends time hunting for answers to recurring questions across Slack, email, and your shared drive. An AI-powered search layer over your existing tools turns the "do you know where the X is" message into a quick answer.
Customer support triage. Incoming customer queries are being read, classified, and routed by hand. AI classifies and routes the routine ninety percent. Your team handles the ten percent that needs real human judgement.
Meeting summaries and action items. Your team is taking notes, writing up summaries, and chasing action items by hand. AI does the capture and structuring. Your team gets time back.
Sales prospect research. Your business development team is reading company websites, LinkedIn profiles, and news articles to qualify leads. AI does the research pass. Your team does the qualification call.
SOP and process documentation. Your team is documenting business processes by writing them up from memory. AI tools record the workflow as it happens and produce a structured draft. We covered this in detail in our post on how to document business processes.
This is not an exhaustive list. It is the list of the most common, highest-leverage first moves we see actually work in growing Australian businesses. If your business has any of these patterns, you have a candidate first move.
How to pick the right one in 60 minutes
You do not need a six-week strategic exercise to pick your first AI use case. You need an hour, your operations leadership, and a willingness to be specific.
Here is the practical exercise.
Step 1. List the top ten recurring operational tasks across your team. Each one in one sentence. Do not overthink it. Write what people on your team actually do most often.
Step 2. For each task, answer three questions.
- Roughly how many hours per week does this take across the team?
- How much human judgement does it actually require?
- If we did this faster or more consistently, what is the downstream effect?
Step 3. Score each one out of three for each question. High frequency, high cost, low judgement requirement scores well. The task that scores highest across all three is your first move.
That is it. The exercise takes an hour with the right people in the room. The output is one clear first move with a specific quantified problem behind it.
The reason this works is that it forces you to start with your business, not with the technology. By the time you have done this exercise, the question of which AI tool to use becomes downstream and much easier to answer. You are looking for a tool that solves the specific problem you have just defined, not a tool you are trying to find a use for.
Running the first 30 days
Once you have picked your first use case, the first thirty days should follow a specific structure to give you the best chance of success.
Week 1. Measure the baseline. Before you change anything, measure how the task is being done today. How long does it take. How often does it go wrong. What is the cost. This is the number you will measure against later, so be specific.
Week 2. Design the new workflow. Map how the task will work with AI in it. Where does the AI fit. What does it produce. Where does a human review or approve its output. What happens when something goes wrong. This is the operational design work that determines whether the AI delivers value or just sits in the workflow doing nothing useful.
Week 3. Run the implementation. Pick the right tool for the task, set it up, train the team on the new workflow, and start running it. Keep the scope narrow. One task. One team. One defined window.
Week 4. Measure and decide. Compare the new numbers to the baseline. Did the AI deliver. Are there issues with the workflow that need refining. Is the team using it the way you designed it. Based on the evidence, decide whether to expand, iterate, or kill the project.
Done well, you have a working AI implementation, a quantified outcome, and a team that has the muscle memory to roll out the next one. Done badly, you have wasted thirty days and a subscription fee, and you know definitively that this was not the right first move. Both outcomes are better than the alternative, which is sitting in analysis paralysis for another quarter.
What to do after the first 30 days work
If your first move worked, you have the basis for everything that follows.
The right next move is not "now let's do AI everywhere." It is "what is the next task on the list that scores high on the same three filters." You go down the list. You implement the second one. Then the third. Each one is faster than the last because your team has now done this before.
Within six months of starting properly, you should have three or four working AI implementations, each delivering measurable value, each owned by someone, each integrated into how the business actually runs. That is a transformed operational layer. It is also achievable with discipline, where most businesses are still on their first stalled pilot.
The mistake to avoid here is scope creep. The temptation, after the first win, is to start expanding the project to include adjacent use cases that seem related. Resist it. Each new use case gets its own analysis, its own design, its own pilot, its own measurement. The discipline of doing one well, then the next one well, is what compounds.
The bigger picture
The reason most operations leaders are stuck on AI is not that the technology is too complicated, the budget is too small, or the team is not ready. It is that they are looking at the entire AI landscape and trying to figure out a comprehensive strategy when what they need is a single first move.
You do not need a strategy. You need to ship one thing that works.
Pick the highest-frequency, highest-cost, lowest-judgement task on your operations team's plate. Quantify the cost of it. Design the workflow with AI in it. Run a thirty-day implementation with a baseline and a success criterion. Measure honestly. Then do it again with the next task.
Done this way, AI stops being an abstract source of pressure and becomes an operational tool that delivers measurable value, every month, in a way the rest of the business can see and trust.
The first move is the only one that matters. Get that one right and everything else gets easier. Skip it, or pick the wrong one, and you will be in the same position next quarter as you are right now.
The right time to make that first move is this month.
