There is no shortage of information about AI.
Every week there is another tool, another model and another example of something AI can do. Businesses are being told they need an AI strategy, employees are experimenting with ChatGPT and Copilot, and there is a growing sense that if you're not doing something with AI, you're falling behind.
But for many business owners and managers, there is a much more practical question:
Where do I actually start?
Recently, I delivered a presentation for CPA Australia on using AI and automation to build a more efficient and scalable business.
One of the main points I wanted to make was that you don't need to start with AI. Start with the business.
Start by looking for friction
Think about how work moves through your business today.
Where are people:
- entering the same information more than once?
- manually reading and processing emails or documents?
- chasing approvals or following people up?
- reconciling spreadsheets?
- preparing the same reports repeatedly?
- searching for information before they can make a decision?
- relying on someone to remember what needs to happen next?
These are the areas I describe as operational friction - re-keying, chasing, waiting and fixing that consumes time without necessarily creating additional value.
As a business grows, that friction becomes increasingly important.
If every new customer means more administration, more handoffs and more dependence on key people, growth becomes expensive. Scale comes from making good work repeatable, visible and easier to deliver. And this is where AI and automation become interesting.
What is AI actually good at?
AI is particularly good at working with information.
It can extract, classify, summarise, generate and recommend. It can process large amounts of information, recognise patterns and help people get to an outcome faster.
Imagine an employee receives an email with a purchase order and several attachments.
Today, they might need to read the email, identify the customer, open the PO, enter the details into another system, create a job, save the attachments and then notify someone that the work is ready.
AI can help interpret all of that unstructured information. But AI alone doesn't know what your business should do with it.
It needs the business context - your customers, jobs, products, rules, documents and current information - and it needs to be connected to a workflow that determines what happens next. That's the difference between using an AI tool and using AI to improve a business process.
So, where do you start?
I use a simple four-step framework:
1. Get the data
What information does the process rely on? Where does it live? Is it accurate? Can you access it? Is there a dependable source of truth?
2. Identify the opportunity
Look for repetitive, high-volume work and information bottlenecks. Don't start with “Where can we put AI?” Start with “Where are we spending time doing work that should be easier?”
3. Define the boundaries
Decide what AI should and shouldn't do. It might prepare something for review. It might automatically process straightforward cases but escalate exceptions. Or it might make a recommendation while leaving the final decision with a person.
4. Connect AI to the workflow
Finally, connect AI to the relevant information and systems so it becomes part of the process — rather than another tool somebody has to manually use. That sequence matters. Technology can't repair an undefined process or unreliable information.
And sometimes the answer isn't AI
This is probably one of the most important things to recognise. Once you've found an inefficient process, AI is only one possible solution.
You might simply remove a step.
You might standardise the process.
You might integrate two systems so information doesn't need to be entered twice.
You might use traditional automation for predictable rules.
You might use AI where information needs to be interpreted.
Or the business might need a custom system to bring the entire workflow together. The objective isn't to use more AI. The objective is to build a better business.
Start with one problem
If you're wondering where to start with AI, don't begin by trying to develop an organisation-wide AI strategy. Find one process. Find something repetitive. Something slow. Something your team complains about. Something that becomes harder every time the business grows. Understand what information it needs, how the process works today and what a better outcome would look like.
Then ask:
Could we remove it, simplify it, automate it - or use AI to make it substantially better?
That's a much easier place to start.
And, in my experience, it's also where the most useful AI conversations begin.
