I have written about cash flow management several times over the past decade.
The reality is simple: you cannot build wealth unless you spend less than you earn and invest the difference (It is the first element in the wealth formula). That is why cash flow management is probably the most fundamental discipline to master if you want to build wealth.
In the past, I have said that the first step towards better cash flow discipline is to measure your cash flow. That means understanding how much you spend and where your money goes.
However, that can be a tedious exercise. So, in this blog, I want to share some important enhancements that can dramatically improve how you manage cash flow, without making the process unnecessarily complicated.
The old method gave you a snapshot, not a picture
For years, I have told people to follow the same process. Download three months of transactions, sort them by description, allocate each line into a handful of categories, total the columns and calculate your surplus.
Three months was a deliberate choice. It is long enough to smooth out minor anomalies, but not so long that the exercise becomes overwhelming. But three months was always a compromise, and that compromise was driven by the effort involved.
Categorising three months of transactions by hand is already a tedious exercise. Categorising two or three years of transactions was simply unrealistic, so almost nobody did it.
That compromise came at a cost. Three months is only a snapshot. It tells you what you spent in one quarter, but it cannot tell you whether that quarter was typical. It misses the natural annual rhythm of spending: holidays, school fees, insurance renewals, car servicing and other expenses that only arise once or twice a year.
It also cannot show you a trend, because trends require time.
When I ask people how much they spend each year on general living expenses, I almost always get a guess. And when households finally sit down and do the numbers properly, more than nine out of ten are surprised by the result.
Feeding years of data to AI changes what you can see
This is where the first important change comes in. I now use Claude, an AI tool, to categorise transactions for me. It does in minutes what could take up to a few hours. That means the practical limit on how much transactional data you can analyse has effectively disappeared.
That is more important than it might first seem.
When you can analyse two or three years of transactions instead of only three months, you stop looking at a snapshot and start seeing the real shape of your spending behaviour.
You can see consistency. Is your discretionary spending relatively stable from quarter to quarter, or does it spike and fall? This matters because consistency tells you how reliable your surplus really is.
You can also see drift. A spending category that increases by 5% per year might be almost invisible over one quarter, but it becomes more obvious over three years. That gradual creep is exactly how lifestyle inflation works. It happens quietly, one small upgrade at a time, until the surplus that used to exist has disappeared.
Most importantly, you can see where the money genuinely goes because you can ask Claude to prepare charts, insights and analysis.
In my experience, most people have a handful of categories that account for most of their potential discretionary overspending. Therefore, my prompt asks you to identify those categories first, so that they can be tracked separately.
Everything else is then allocated into the same seven categories I have always used: three non-discretionary categories and four discretionary categories.
Claude will also audit the data you provide and produce a list of transactions it cannot categorise with confidence. You can then review those transactions, clarify how they should be treated, and feed that information back into Claude.
Once you have done that, Claude can produce a very useful analysis of your spending patterns.
The final step is also important. After the analysis is complete, Claude will offer to write a refined prompt that you can use next time you repeat the exercise. This means it can progressively learn more about your transaction history, your spending habits and your preferred categorisation rules. In other words, the process improves each time you use it. If my prompt below doesn’t do this automatically, just write “I’m happy with the analysis. Please now generate the Pass 4 personalised prompt, as described in my original instructions.”
>> Download the prompt here <<
A word of caution before you celebrate
Claude is a fast and capable assistant, but it is not a substitute for judgement. There are two things it still cannot do for you.
First, it cannot know which transactions are genuine one-off expenses. Once everything has been categorised, you still need to identify non-recurring items, such as a major repair or another unusual expense, and exclude them from your maintainable spending figure.
However, be careful not to exclude costs just because they are irregular. If unusual expenses keep appearing throughout the year, they are probably not unusual. For example, if an old car breaks down every few months, it would be sensible to include a modest annual allowance for repairs. Claude cannot make that judgement reliably because it requires knowledge of your life that does not exist in the transaction data.
Second, your bank data is sensitive. Before uploading anything, remove account numbers, names, BSBs and any other identifying information. Claude only needs three pieces of information: the date, the transaction description and the amount.
Also, be aware that AI tools differ in how they treat the information you upload. Some may use your inputs to improve their models unless you change the relevant setting. Check this before you start rather than making assumptions either way.
Knowing your number is only half the job
The truth is that better analysis alone does not solve. Insight without structure rarely changes behaviour.
You can produce a beautiful multi-year breakdown of your spending and still fail to build wealth, because knowing what you spend and consistently spending less are two very different things.
The first is a measurement problem – you cannot manage what you don’t measure. AI has largely solved that.
The second is a behaviour problem. And behaviour is where most people come undone. The reason is simple. Willpower is unreliable. It’s easy to buy things online. Tap your card. So, the goal is not to rely on discipline. The goal is to design a system that does not require much discipline in the first place.
Automation is the real key to cash flow management
The most effective cash flow management is the kind you do not have to think about, because it has been automated.
A well-designed banking structure makes good decisions automatic and bad decisions inconvenient. It does this through three principles. All three can be set up once and then left to run.
First, pay yourself first.
As soon as your income arrives, a set amount should be directed towards saving, investing or debt reduction before anything else happens. Treat it like a non-negotiable bill.
The key is that you do not make this decision every pay cycle. You make it once. You set up the automatic transfer or direct debit, and then it happens whether you remember or not.
Second, isolate your discretionary spending.
A fixed amount should automatically move into a separate spending account at the start of each cycle. All discretionary expenses then come from that account.
When the account is empty, you stop spending.
There is no willpower involved, because the structure imposes the limit for you.
Third, separate your savings from your spending.
Keep your long-term money out of sight, ideally hidden from your everyday banking view. If you cannot see it, you are less likely to casually spend it.
Many employers will split your salary across multiple accounts. That means your pay-yourself-first amount can go straight to savings, an investment account or your loan, and never pass through your everyday spending account at all.
The beauty of this approach is that the difficult decision, how much to save and how much to spend, is made once, when you set the transfer amounts.
After that, the system quietly does the heavy lifting.
You are no longer relying on being disciplined every week. You are relying on a structure that makes the right behaviour the default.
Bringing the two halves together
These two shifts work best as a pair.
The AI analysis gives you an accurate, evidence-based view of what you actually spend. It is drawn from years of data, not a single quarter. That helps you identify your true maintainable spending and, by extension, your real surplus.
The automated banking structure then takes that surplus and puts it to work without relying on your ongoing effort.
In other words, the first step tells you the right numbers i.e., how to allocate your income. The second step turns that number into a system.
You use the analysis to decide how much should be transferred automatically towards saving, investing or debt reduction. Once those transfer amounts are set, the system runs in the background.
You should still review the analysis periodically, ideally every six to twelve months, or sooner if your circumstances change. Incomes rise, expenses shift and a system that suited you two years ago may no longer be appropriate.
But now that the analysis takes minutes rather than hours, there is no real excuse not to review it.
A practical takeaway
For a long time, the barrier to good cash flow management was effort.
Working out what you really spend was tedious. Acting on that information required constant discipline.
Both barriers have now been lowered.
You can analyse years of your own spending in minutes and identify your maintainable spending from actual behaviour, not guesswork. You can then build a banking structure that directs your surplus towards saving, investing or debt reduction automatically.
That means building wealth no longer depends on being disciplined. Measure, analyse and identify where you can make improvements. Then automate it.
