There is a dangerous sort of comfort that comes from having a good spreadsheet.
Mine had tabs, formulas, colour coding, target allocations, a gilt ladder, investment categories, and a dashboard view. It looked disciplined and felt controlled.
And then I realised something awkward:
My portfolio was not quite what I thought it was.
Not because the numbers were wrong. The numbers were fine.
The problem was how I had been thinking about them.
This post is about that journey: using a spreadsheet, some uncomfortable questioning, and AI assistance to get from “I have a plan” to “I understand what the plan is actually doing”.
This is not financial advice, and it is not a recommendation to buy or sell any investment. It is a personal learning note about how I use a spreadsheet, cashflow modelling and AI-assisted questioning to understand my own retirement plan more clearly.
But it may help anyone approaching retirement, or anyone trying to make their investment spreadsheet more honest.
The first mistake: treating “play money” as if it wasn’t real risk
I had been thinking about my investments in separate pots:
- global equity
- gilts
- cash
- AVCs
- ex-US equity
- a smaller “play” portfolio of individual holdings
Emotionally, the play portfolio felt separate. It was the sandbox. The bit where I could take more risk, learn, experiment, and satisfy the natural human urge to fiddle.
But mathematically, it was not separate at all.
It was equity risk. In fact, it was probably higher-risk equity exposure.
That was the first big insight.
I had been thinking I was moving towards a sensible retirement allocation. But once I counted the play portfolio properly, my equity-like exposure was higher than I realised. Not absurdly high for someone with a long-term horizon, perhaps, but too high for the version of the plan I thought I was running.
The lesson:
If it can fall like equity, count it as equity.
A “play” portfolio is not outside the risk budget just because it has a different label.
The cleaner way to think about allocation
I now prefer a much simpler top-level split:
Growth assets
These include:
- global equity trackers
- AVCs invested in equity funds
- ex-US equity tilts
- individual shares
- small-cap or speculative holdings
- the play portfolio
This also links to my broader view that passive investing is enough for most people, provided the overall structure is honest.
Defensive assets
These include:
- cash
- gilts held to maturity
- money market-type holdings
- anything intended to fund spending without relying on equity markets
That gives a much clearer question:
What percentage of the portfolio is genuinely exposed to growth risk, and what percentage is genuinely defensive?
For me, the useful direction of travel became something closer to a true 60/40 structure, where the 60% includes all equity-like risk, not just the neat index-tracker part.
That distinction matters.
A portfolio that is “60% global equity plus 10% play” is not 60/40. It is much closer to 70/30.
Obvious, perhaps. But only once you look at it properly.
The second mistake: looking at the gilt ladder annually
A gilt ladder is meant to reduce sequence-of-returns risk. The idea is that gilts mature over time, producing known cashflows, so you are not forced to sell equities during a downturn.
Initially, I looked at my ladder by year.
That is useful, but it is not enough.
Retirement spending does not happen annually. It happens monthly.
A year might look “covered” if enough gilts mature during it, but that does not tell you whether the cash arrives before you need it.
For example, if a gilt matures late in the year, it does not help much with spending in the spring and summer unless you have cash to bridge the gap.
So the better question became:
If I draw a fixed amount every month, when does the running balance go negative?
That was a much more useful lens.
Turning the gilt ladder into a cashflow model
This is where AI was genuinely helpful.
I already had a table of gilt holdings. What I needed was not another table of holdings. I needed a cashflow view.
So I moved towards a structure like this:
| Month | Maturity cashflow | Monthly spending need | Running balance |
That changed everything.
Instead of saying:
“How much matures in 2029?”
I could ask:
“Does the ladder keep the running balance above zero if I spend £X per month?”
That exposed timing gaps.
It also made potential purchases much easier to judge. A new gilt purchase was not attractive simply because the yield looked good. It was attractive if it improved the running balance at the point where the plan was weak.
That is a much better decision test.
The third mistake: confusing yield attraction with planning need
When gilt yields rise, it is tempting to think:
“I should lock that in.”
Sometimes that is sensible. But it can also be yield chasing.
The distinction I now use is this:
Good reason to buy gilts
“This maturity fills a known spending gap in the ladder.”
Weak reason to buy gilts
“The yield looks attractive and markets feel uncertain.”
The first is planning. The second is market timing wearing a sensible coat.
This became especially important in the current environment, where inflation has been sticky and yields look tempting. A gilt yielding more than 4% can look very appealing when markets are volatile. But that does not automatically mean selling equities to buy gilts is wise.
The better question is:
What job is this gilt doing?
If the answer is “funding a known gap”, that is a solid reason.
If the answer is “making me feel safer this week”, that needs more scrutiny.
The ex-US debate: useful hedge or another thing to fiddle with?
Another part of the journey was thinking about global equity concentration.
Most all-world trackers are heavily weighted towards the US, and particularly towards a small number of very large technology companies. That concentration may be justified. It may also be a vulnerability.
I considered adding a developed ex-US allocation as a modest counterweight.
The key was to define its purpose clearly.
It was not a crash hedge. In a broad sell-off, global markets often fall together.
It was not a bet that Europe or Japan – or even China – would suddenly dominate the world.
It was simply a way to reduce reliance on US mega-cap dominance.
That led me to a modest allocation rather than a dramatic one. Enough to acknowledge the concentration risk. Not enough to create a new obsession.
The lesson:
A small tilt can be sensible, but only if you know what it is for.
If you cannot explain the role of a fund in one sentence, it probably should not be in the portfolio.
The ETF plumbing lesson
I also learned a practical lesson about global ETFs.
A fund can look excellent on headline fee and exposure, but still be suboptimal for your setup.
In my case, I had moved into a very cheap global tracker. The exposure was broadly right. The fee was low. But it was distributing rather than accumulating, and the cashflows created awkward foreign exchange friction.
The problem was not currency exposure in the underlying investments. Any global fund will have that.
The problem was operational friction: dividends being paid out, converted, and then needing to be reinvested.
That is not catastrophic, but it is annoying and potentially costly over time.
So the conclusion was not “chase the cheapest ETF”. It was:
Choose the cleanest ETF for your actual platform, tax wrapper, and behaviour.
For me, an accumulating global ETF traded in sterling is cleaner, even if the headline fee is a little higher.
The fee difference is noise. The operational simplicity is valuable.
How AI helped
I did not use AI to outsource the decision. That would be dangerous.
I used it as a thinking partner.
This is also why I think “AI” may be the wrong name for what these systems are becoming.
The most useful thing was not that it gave answers. It challenged the framing.
It asked, in effect:
- Are you hedging or making a bet?
- Is this a permanent allocation or a reaction to headlines?
- Is your play portfolio really outside the risk budget?
- Are you buying gilts to match liabilities or because yields look good?
- What happens if markets rise after you de-risk?
- What happens if markets fall and you hesitate to re-enter?
- Are you solving the right problem?
That sort of challenge is useful because most investing mistakes do not begin with bad arithmetic. They begin with sloppy definitions.
AI also helped with practical shortcuts:
- designing helper columns
- building pivot-table ideas
- separating calendar year, fiscal year, and funding year
- creating monthly cashflow views
- modelling running balances
- testing different monthly spending assumptions
- generating charts from table data
- turning a messy conversation into a clearer plan
In other words, AI helped turn the spreadsheet from a record of investments into a model of future cashflows and that made a big difference.
The most useful spreadsheet shortcut
The single most useful change was this:
Stop only tracking holdings. Start tracking cashflows.
For each gilt, the key planning data is not just:
- ticker
- price
- yield
- maturity date
It is:
- when the cash arrives
- how much arrives
- what spending period it is intended to fund
- what happens to the running balance afterwards
A simple chart showing cashflow against monthly spending need made the ladder much easier to understand.
It showed where the plan was strong.
It showed where it was weak.
It showed which maturity would actually help.
That is far better than staring at a list of yields.
The big lesson: build around the job to be done
The most important question is not:
Which investment looks best?
It is:
What job does this part of the portfolio need to do?
For my plan:
- global equities provide long-term growth
- the ex-US sleeve modestly reduces US concentration
- gilts provide known future cashflows
- cash bridges timing gaps
- the play portfolio satisfies curiosity but must stay inside the risk budget
- the spreadsheet keeps the whole thing honest
That is the structure.
The aim is not to predict markets. It is to avoid being forced into bad decisions when markets are unhelpful.
Where I have landed
The portfolio I thought I had was not exactly the portfolio I actually had.
That was uncomfortable, but useful.
The direction now is clearer:
- count all equity-like exposure honestly
- keep the play portfolio capped
- move towards a more deliberate growth/defensive split
- use gilts to match spending needs, not to chase yield
- keep global equity simple
- avoid unnecessary ETF tinkering
- model cashflows monthly, not just annually
- let the spreadsheet drive decisions, not headlines
That last point matters most.
A good retirement plan should reduce the number of decisions you have to make under stress.
The spreadsheet should not just tell you what you own.
It should tell you what happens next.
If you like practical tools and systems for making life a bit less over-engineered, you can also get my free guide: 10 Tools That Help Me Earn Smarter, Move Faster, And Live Freer.
Over to you
I’m still refining this model, and part of the reason for sharing the journey is to make the learning visible rather than pretending I had the answer from the start.
I’d welcome feedback, challenge, or questions.
In particular, I’d be interested to know where a deeper dive would be useful. For example:
- a beginner’s guide to gilt ladders
- how to think about sequence-of-returns risk
- how to classify portfolio risk properly
- how to use AI as a spreadsheet/planning assistant
- how to build a simple monthly retirement cashflow model
- the difference between calendar year, tax year, and “funding year”
- how to avoid over-optimising ETFs
- how to keep a “play” portfolio without letting it distort the whole plan
If any of those would be useful, or if there’s another part of the journey you’d like unpacked more clearly, let me know. I’m learning as I go, and I suspect many of us are wrestling with the same questions in slightly different spreadsheets.
