CFOO Says the Real AI Project Is Clean Data, Not Clever Automation
Dave Leaver explains why clean data is the foundation of AI in finance and how automation can free CFOs to focus on judgment, strategy, and decisions.
Dave Leaver
CFOO of Mention Me
Faster Close: AI reduced month-end close from ten days to three while improving board reporting accuracy and relevance.
Clear Roles: AI handles calculations, assembly, and pattern detection; humans retain accountability for assumptions, risks, capital allocation, and recommendations.
Data First: Reliable connected data matters more than sophisticated models, because poor foundations produce polished outputs leaders cannot trust.
Dynamic Hiring: Driver-based headcount planning links hiring triggers to revenue, runway, and cash, enabling faster, evidence-based staffing decisions.
Practical Adoption: CFOs should test AI on real finance work now, rather than waiting for tool evaluations or perfect products.
Dave Leaver is the CFOO of Mention Me, a Series B B2B SaaS business in the referral and advocacy space.
We sat down with Dave to learn where AI provides the greatest leverage in finance workflows. Here's what he said.
CFOs Should Be Able to Set Direction — Not Just Report on It
I'm Dave. I'm CFOO at Mention Me, a B2B SaaS business in London. I'm also a co-founder of HP-1, a creator-led DTC brand I'm building from the ground up.
Create a Free Account to Read More
You'll also get access to a growing community of modern CFOs and finance executives accessing proven frameworks, tools, and insights to navigate AI-driven finance.
Have an account? Log In
My route wasn't conventional. I started out chasing a career in professional sport. When that didn't happen, I trained as an accountant — then moved to Sydney to chase the sun.
Twenty-plus years on, I've worked across consulting at KPMG, FMCG and ASX-listed businesses, run finance for celebrities, and built deep experience in VC- and PE-backed tech startups on both sides of the world. I've built finance functions from scratch, scaled companies across continents, and owned P&Ls worth hundreds of millions.
But I never fit the stuffy accountant stereotype. I love simplicity, and I think work must be impact-driven. If a task isn't moving something forward, I eliminate it.
That instinct put me at odds with how finance traditionally operates. Thorough, accurate, and a step behind the business — so the numbers are right, but by the time the analysis lands, decisions are already made.
So, when AI came along, I embraced it — not as a trend to adopt, but as a way to strip out manual work. I spend less time reporting numbers, and more time on what they mean and where the business should go next. The same systems I use to run my own companies give me real-time numbers, board packs that tell a story rather than just report one, and a lean finance function that scales without increasing headcount.
I believe a CFO should be able to set direction, not just report on it after the fact.
A Lean Team Powered by AI
Mention Me is a Series B B2B SaaS business in the referral and advocacy space, headquartered in London and backed by Octopus Ventures and Eight Roads. We sell to brands across the UK, Europe, and beyond.
My remit is broader than finance alone. It spans finance, HR and people, strategy and operations — so I'm as close to how we hire and organize as I am to the numbers.
The structure is lean by design. A small in-house team handles the core, with an outsourced function supporting transactional work. The function runs on AI. We've automated reporting, data pipelines, and first-pass analysis — work that previously required more hands. This allows the team to spend its time on judgment rather than production.
I've taken the same approach to my own role. I've built an AI "chief of staff" that helps me run day-to-day, triaging my inbox, flagging what needs a reply, capturing actions out of meetings, and pulling it all into a morning brief so I start each day knowing exactly where to focus.
Where AI Excels Versus Where Humans Are Necessary
AI does the production, humans do the judgment — mostly.
On the AI side sits anything involving patterns, calculations, or assembly. This includes the number-crunching in forecasting, building the model, rolling actuals forward, and flexing scenarios. For variance analysis, AI excels at spotting movement and identifying its source. It also handles cash flow modeling, the mechanics of the board pack, and first-pass commentary. Essentially, this is the work that used to eat the week. AI is faster than me, it doesn't get tired, and it doesn't make arithmetic mistakes at 9 pm.
Humans explicitly handle everything that carries consequences or requires context the machine lacks. Capital allocation — deciding where the next pound goes — is a judgment call about strategy and risk appetite, not a calculation.
The same applies to the assumptions underneath any forecast: AI can run the model, but deciding whether we really will land that deal or hold that price is a human call, and the output is only as good as that judgment. Risk assessment, anything involving people, and anything I'm putting in front of the board with my name on it — that's mine.
The reason for the line is simple. AI is brilliant at telling you "what" and "where." It's much weaker at "why", and it has no skin in the game on "so what do we do about it."
Variance analysis is the clearest example: AI tells me the number moved and points to the driver in seconds. What we do next — that's the part I'm paid for.
How Month-End Close Can Be Cut From 10 days to 3 Days
AI handles production end-to-end, right up to the point where consequences begin. That’s where I step in.
I've recently decreased the time it takes to complete month-end close from ten days down to three.
Before, I had to pull the numbers, reconcile, build the slides, write the commentary, rebuild the charts — most of it by hand, with one or two people handling the majority. By the time the board pack landed, it described a position weeks out of date.
I automated the production and kept the humans on the judgment. We now close in three days — including the board pack and updated forecasts.
Here's the process, from raw ledger to a board-ready pack.
- The process starts with data. Instead of exporting figures and pasting them into a spreadsheet, the platform pulls live data from source systems — the ledger, payroll, sales pipeline, and subscription metrics — into one connected model. This single step removes most manual effort and copy-paste errors.
- It structures and checks the data: It standardizes the numbers, automatically flags reconciliation issues, and updates the forecast from actuals in the same pass, ensuring that the forward view never lags the reported one.
- It builds the outputs. The charts and board pack are assembled from that one source, and AI drafts first-pass commentary on what moved, so I start with a draft instead of a blank page.
- The human takes over — the part that matters. I pressure-test the narrative and decide what the board needs to hear: what's a blip, what's structural, and what decision I recommend. The machine provides a strong draft in hours; the judgment about its meaning remains mine.
AI handles production end-to-end, right up to the point where consequences begin. That's where I step in.
Speed and quality improved at the same time. The pack reflects where the business is now, not three weeks ago. Because I no longer wrestle the numbers into shape, I dedicate time to shaping the narrative, understanding what drives the movement, and determining which decisions to present to the board. In short, we stopped reporting the numbers and started explaining them.
The Pros and Cons of AI in Finance
The good results are measurable. Beyond the month-end close improvements, scenario planning now takes minutes instead of a day or two. And automating production work has saved over six figures in finance headcount costs that the function would otherwise have needed — a cost saving that pays for everything else.
A less obvious, qualitative win is the time I've saved. The hours I used to spend assembling numbers now go into interpreting them, which makes board conversations sharper and transforms the function from a reporting function into a decision-making one.
But a real downside exists. AI is sometimes confidently wrong. It provides polished-looking answers that are subtly off — a misread assumption, a number that doesn't tie out. Its fluency makes it easy to trust without checking.
Ask it for life advice, and you'll see the same thing in a more entertaining form: supremely confident, it normally agrees with you, and it's often complete nonsense.
The finance version is just harder to spot. Early on, it bit me more than once. I've had to build and instill in the team the discipline to treat AI output as a first draft, never a final answer. A human still checks every number presented to the board.
Why CFOs Must Set the Groundwork Before Integrating AI
The data foundation is the project, not the warm-up to it.
When I started, I thought the hard, valuable work was the AI: the models, the automation, the clever output. I treated cleaning and connecting the data as quick prep before the real build began. That was exactly backward.
The groundwork was the build. Most of the effort, and most of what determined whether the thing actually worked, lived in the unglamorous layer: clean data, sensible structure, and reliable connections to source systems.
As a result, I spent early effort building cleverness on top of unready data, and got polished output I couldn't fully trust. This meant going back and doing the foundational work anyway, after the fact. Had I known, I'd have sequenced it the other way: Get the plumbing right first, then layer the intelligence on top. Same destination, half the wasted motion.
The magic only works once the boring groundwork is done.
How CFOs Can Enhance Headcount Planning with AI
In most businesses, and certainly in SaaS, where people are by far the biggest cost, headcount planning remains a blunt instrument. These businesses use a spreadsheet of roles, a hiring plan set at budget time, and a number everyone treats as fixed until something forces a rethink. This is strange because it's the single largest lever a CFO controls, yet they often model it in the least dynamic way.
I redesigned our headcount-planning approach, shifting from a static list to a driver-based model wired into the rest of the plan. Rather than "we'll hire twelve people this year," it's "this role is triggered by this much pipeline, this one by this revenue threshold..." so hiring connects to the things that justify it.
Because it lives inside the connected model, I can flex it live: add a hire, change the timing, and watch it roll straight through the P&L, cash position, and runway.
As a result, headcount stopped being an annual guess and became a real-time decision. When someone asks, "Can we afford this hire?" I can answer it in the meeting, not with a gut feel, but with its impact on runway and timing. This turns one of the most consequential and emotive conversations in the business into one grounded in numbers, on the spot.
Why Claude Is a Must-Have Tool for CFOs
Claude is my go-to tool, without hesitation. It's not a single-purpose finance tool, and that's exactly why I love it; it handles everything other tools cannot flex to.
Honestly, I expected more from off-the-shelf finance tools that have bolted "AI" onto an existing product. Most of them are chat boxes on top of the same software, and they don't meaningfully change how the work gets done. They do one job well — if that.
But Claude can do it all. It builds the models, drafts the board commentary, interrogates the data, writes the analysis, and acts as the thinking partner I bounce decisions off. It handles the wide range of tasks that fill my week and adapts to whatever I throw at it next.
Here’s my advice: The biggest mistake I see is CFOs treating AI as a procurement decision — a twelve-month evaluation, pilot, and rollout. Don’t wait for the perfect tool. Most importantly, get the boring bit right. And keep your judgment central. AI changes how work gets produced, not who’s accountable for it…
The Biggest Mistake That CFOs Make with AI
Here's my advice:
- The biggest mistake I see is CFOs treating AI as a procurement decision — a twelve-month evaluation, pilot, and rollout. I've made that mistake. By the time that process finishes, the tools have changed twice. Instead, use it yourself, on real work, this week. You'll learn more about where it helps and where it lies to you in a fortnight of hands-on use than from any vendor demo.
- Don't wait for the perfect tool. The off-the-shelf market still mostly offers a chat box bolted onto old software. You can build more than you think; the barrier to creating your own solutions has collapsed. CFOs who'll pull ahead will treat that as a capability to develop rather than a service to buy.
- Most importantly, get the boring bit right. Clean data and sound process aren't just an unglamorous prerequisite to AI; they're the whole game. AI on top of messy data just produces confident nonsense faster. Fix the foundations first.
- And keep your judgment central. AI changes how work gets produced, not who's accountable for it. The skill that matters more than ever isn't producing the analysis — it's knowing whether it's right and what to do about it. That's still the job. It always was.
Follow Along
You can follow Dave Leaver's work on LinkedIn.
More expert interviews to come on The CFO Club!
