I spent a large part of my corporate career standing in front of a whiteboard with a black (sometimes red) Texta in my hand.
Before Excel was opened. Before formulas were written. Before tabs were colour-coded.
I would sketch the thoughts of the business first.
What we thought volumes would do. What the demand signal looked like for the next quarter. What the market was telling us. Impact on margins. Pricing. Mix. Any timing and moving stuff around.
Only after that would I go away and build the financial model, run the algorithm, dump the data out, and play around.
The unspoken rule was simple.
If the numbers didn’t line up with the whiteboard, I needed to be able to explain why. Not technically. Not defensively. Commercially.
Because most executives already have a model in their head.
They carry pattern recognition. They carry years of experience. They carry an instinct for the company and the market that has been honed by walking the floor, talking to consumers, watching competitors, and sitting with customers.
They simplify the business quickly. They reduce a lot of complexity to a small number of moving parts.
And when finance presents a forecast that doesn’t “feel right”, credibility quickly collapses and confidence drifts. When confidence drifts, influence follows. And when influence is gone, you become the team that produces information nobody acts on.
Over time, I realised financial models are not valuable because they are precise. They are valuable because they are trusted. They are valuable because they help leaders act.
Now this isn’t a be-all and end-all list. But here are five things I believe every financial model must consider, plus a sixth I would not have written a year ago.
Not because they make the model smarter. Because they make the conversation better.
1. Address all risks and opportunities
Many financial models still present a single outcome. One revenue number. One cost expectation. One profit line. One EBIT — usually a specific number that has to be right because it’s so specific.
But executives look at it, and the first thing they think is: what’s in and what’s out.
They want to know what could go wrong, what might go better, and which assumptions are most fragile.
Your forecast means nothing without R&Os – risks and opportunities – being clear to all.
What’s in. What’s out.
That instantly tells executives where to take the conversation.
A number bang on budget that has all of the opportunities in and none of the risks is madness. A number 90% short with all of the risks loaded in and none of the opportunities is finance conservatism. You must show both so the leadership team gets the feel of the range, not just the centre point.
This is essentially scenario planning, although most people don’t dress it up with that label. A bear case. A base case. A bull case. With the assumptions visible. What sales volume needs to be held? What pricing needs to land? What the marketing investment is delivering. What consumer behaviour are we betting on? What expense lines could blow?
When you do that, debate improves, ownership increases, and decisions accelerate.
The discussion moves away from defending spreadsheets and towards shaping direction. Accountability shifts from finance to the leaders who actually pull the levers.
That shift is subtle, but powerful.
2. Include a year ago comparison
Budget comparisons dominate most performance conversations. Yet budgets are negotiated outcomes.
They reflect ambition. They reflect compromise. They reflect what we want the number to be.
A Year Ago comparison often provides something more grounded. Perspective. Context. Visibility into what actually happened the last time we ran this play.
Credibility. Why? Because Year Ago (YAGO) actually happened. It’s not a spreadsheet.
Executives instinctively benchmark against lived experience.
Does this growth feel achievable?
Does this cost position feel sustainable?
Does this customer retention rate feel familiar?
Does the churn rate make sense given what we did with pricing last quarter?
Does the cost of credit and the interest line up with what’s happened to rates in the wider market?
Does this trajectory feel familiar?
Year Ago views help answer those questions. They show momentum building or fading. They reveal structural shifts or temporary noise. They highlight whether optimism is supported by evidence or just enthusiasm.
YAGO also strips out the games people play with budgets. The back-loaded forecast was built to appease the board. The hockey stick that always tilts up in Q4. The 5% productivity gain that never quite arrives. The employee turnover assumption that quietly improves on paper, but not in the building.
A budget can be massaged. A year ago can’t.
3. Gap closing
Every organisation operates with targets. Revenue targets. Profit targets. EBIT targets. Cash flow targets. Goals and KPIs for customer retention, productivity, headcount, and the handful of performance indicators that executive management actually tracks on a Tuesday morning.
Performance conversations naturally gravitate to gaps.
Yet many financial models stop at measurement.
They quantify shortfalls. They show variance. They signal concern. Then they pause.
Your model is the start of the executive-level conversation, not the end of it.
You want to be thinking about which levers could move the outcome. Pricing. Discounting discipline. Marketing reallocation. Customer retention plays. Resource allocation across business units. Timing of expenses or hiring. Which assumptions could be challenged. Which scenarios are worth modelling out fully.
This is not about finance dictating execution. It is about finance shaping possibilities.
When models include potential responses, they become tools for leadership rather than reports for explanation. They start the conversation of “Now what?”
And leadership rarely needs more explanation. It needs more clarity, more choice, and more options to close the gaps.
The best gap-close conversations look almost nothing like a budget review. They look like a strategy meeting with numbers attached.
4. Does it pass the whiteboard test?
Most executives run a simplified commercial model in their mind before the forecast hits their inbox.
They dumb it down to the simplest form in their head before you even present. They have an expectation of what you will show them.
If your model produces outcomes that sit far outside that intuitive view, something important is happening. Either your assumptions are flawed. Or conditions have shifted. Or their mental model is out of date.
So you need to either explain the gap, or risk being picked apart when you say “that’s what the spreadsheet said” – or worse, “that’s what the AI built”.
Leaders rarely challenge numbers to be difficult. They challenge them because the logic does not match the picture in their head.
This is also where transparency saves you. If the numbers come from a single source of truth that’s traceable to the underlying drivers and the data behind them (your CRM, ERP, the SQL view that feeds the warehouse, whatever clean version your finance team has actually built), you can walk the executive line by line through the difference. If it can’t be traced, you’re guessing in a nice font.
If you can’t explain why the model differs from the whiteboard in their mind, you’re lost already.
5. Driver-based forecasting beats mathematics and algorithms
Trend forecasting is becoming increasingly automated. Algorithms extend history. AI detects patterns and produces a prediction.
Systems generate projections. This is useful. This is fast.
But trends change. Customers change behaviour. Capacity moves. Pricing power shifts. Competitors do weird stuff like begin to compete properly.
Driver-led forecasting starts somewhere else.
It focuses on the activity that leads to the outcomes. Examples include:
- FMCG: Forecasting sales based on the promotional calendar by product family, by customer, by week, with expected volume uplift, trade spend, and margin impact.
- Professional services: Forecasting revenue based on billable headcount, utilisation rates, charge-out rates, and weighted client pipeline.
- Property: Forecasting revenue and cash flow based on project milestones, settlement dates, occupancy rates, lease renewals, rental escalations, and development completion timing.
The cash flow falls out of the drivers, not the other way around.
It is operationally undisputed. A driver-based model reflects operational truth in a way that trend extrapolation never can.
Building forecasts from drivers forces deeper engagement with the business. It forces explanation when outcomes diverge from the trend. It forces ownership of assumptions. It brings FP&A closer to decision-making.
And that proximity makes you more useful and more relevant.
Pigment notes that, despite the obvious benefits, only a small share of organisations actually run fully driver-based models. Most still default to top-down growth percentages or last year plus five. That gap is where the credibility lives.
6. Surviving the AI agent
A year ago, I would hand a list like this over and trust people to apply it.
Now I walk into rooms where the FP&A team has an AI agent running in the background while we talk. Pigment Copilot. Anaplan Finance Analyst. Workday Illuminate. Cube. Jedox. Microsoft Copilot for Finance. Or just Claude or ChatGPT plugged into a spreadsheet.
The agent ingests the data. It reads the drivers. It updates the forecast. It drafts the variance commentary. It throws three scenarios back at you before you’ve finished your coffee.
And honestly? A lot of it is impressive.
But it’s also where the five must-haves get tested hardest.
Because an agent will give you a single-point prediction unless you make it carry R&Os. It will benchmark the budget unless you teach it to lean on YAGO. It will quantify the gap and stop there unless you’ve told it to surface levers. It will pattern-match to history unless you’ve forced it to start from drivers. And it will absolutely produce a number that does not pass the whiteboard test, with full confidence, in beautiful natural language.
The risk isn’t that AI replaces the finance business partner. The risk is that AI produces a polished, plausible, fully-formatted forecast that nobody on the executive team actually believes.
Workday’s 2026 FP&A Blueprint calls this year the mainstreaming of agentic AI for planning. Anaplan’s write-up of its Finance Analyst agent describes agents that continuously monitor revenue, expense, margin, and cash flow drivers and propose forecast adjustments without being asked. FP&A Trends makes the same call. The static annual budget — that negotiated, compromised, ambition-laden spreadsheet most organisations still update twice a year — is quietly being retired in favour of rolling, continuous, market-aware forecasting. Some people are calling it xP&A. Whatever you call it, the cycle is shorter, and the cadence is higher.
And it changes what the job actually is.
When the model takes thirty seconds to build instead of three days, the bottleneck moves. It moves from construction to defence.
Defending why the agent’s forecast feels right. Or doesn’t. Defending why churn rate is modelled at 6%, not 4%. Defending why customer retention assumptions hold up. Defending why revenue growth leans on sales productivity, not just trend. Defending which scenario the leadership team should plan against, and what resource allocation actually follows.
The whiteboard test stops being a gut check at the end. It becomes the input.
Here’s how it plays out in practice now.
You still start at the whiteboard. Volumes. Margin. Mix. Pricing. Demand signals. Expected cash flow. Likely employee turnover. Interest rate assumptions. Credit terms. Marketing investment. Whatever the commercial levers are for your business model.
Then you let the agent do the lift. Pull the data, build the driver tree, run the sensitivity, draft the variance commentary, and generate three scenarios with risks and opportunities attached. Most decent setups now connect the agent to a single source of truth — a planning platform, a data warehouse with SQL views into the finance system, whichever version of “one version of the truth” your organisation has actually built. The agent works against that, not against scattered exports and stitched spreadsheets.
Then you do the bit it can’t do.
You ask whether the answer makes commercial sense. You ask which assumptions the executives will challenge. You ask what’s missing — the consumer behaviour shift the data hasn’t caught up with, the competitor move, the negotiation about to land, the credit terms changing, the cost line about to spike. You ask which scenario the business should plan against. You ask what we do.
That’s the human-in-the-loop bit. That’s the bit that puts you in the room.
A few practical things I’d add for anyone running a model alongside an agent right now.
Make the agent show its working. Transparency on which drivers are used and how sensitive the output is to each. If it can’t tell you, you don’t have a forecast – you have a guess.
Don’t accept a single number. Force a range. Force the R&Os.
Cross-check against YAGO before you cross-check against the budget. Year ago is the truth. Budget is a wish.
Audit the assumptions, not the arithmetic. The arithmetic will be right. The judgment might be hallucinated.
Keep the whiteboard. Literally. I still draw the picture before I look at the agent’s output. If the picture and the output disagree, that is where the conversation is.
And keep the accountability with a person. Not the platform. Not the model. Not the agent. A person owns the number when it goes to the board.
A lot of the people I work with in The Leaders Lab arrived because the gap between building a model and defending it in the room had become uncomfortable. If that sounds familiar, check out the training here. It’s the work I wish someone had walked me through when I was the one with the Texta in my hand.
Closing the loop
The five must-haves haven’t changed. The tools have. The complexity has shifted from how you build to how you challenge.
I love financial models. I built a career on them for the first part of my commercial life.
Then I realised they rarely reflected what my years of experience intuitively knew and could shape. And those two things had to align, or finance was considered next to useless.
Finance shouldn’t build models to be admired. They should build them to be used to help shape decisions. Used in debate. Used in planning. Used in moments where judgement matters.
Leadership is rarely about perfect forecasts. It is about informed judgement and timely action that follows.
Whether you use a spreadsheet, Claude, an FP&A platform, or just throw darts at the wall: if you don’t know what’s in and what’s out, if you haven’t built it from drivers, if it doesn’t pass the sniff test, if you can’t bring ideas to close the gaps, and if you’re letting an agent do all of it without challenging the output…
…then your model is just a spreadsheet/AI fantasy nobody can use.