73% of CFO Functions Are Automatable. Most Companies Have Automated Zero.

I've spent the better part of two decades inside finance functions - as a CFO, a fractional CFO, and now as someone who rebuilds them from the outside. And there's one thing I see consistently across every engagement: the gap between what's possible and what's actually happening in finance operations is enormous.

We're not talking about some futuristic capability. The tools to automate the majority of what finance teams do today - reconciliations, variance commentary, close checklists, board reporting, expense processing, revenue forecasting - exist right now. Most of them are affordable. Many are available immediately.

And yet, the median finance team at a $20M–$100M company is doing essentially the same manual work they were doing in 2015. Spreadsheet updates that take 6 hours. Board packages assembled by hand over a weekend. Month-end closes that drag to day 12 or 14. Revenue forecasts rebuilt from scratch every week.

The question I get asked most often is: why?

The automation gap is real - and larger than most CFOs realize

McKinsey's research puts the figure at 73% - the share of CFO function tasks that could be automated with technology available today. Gartner's analysis lands in a similar range. Deloitte's surveys of finance leaders consistently show that most acknowledge the opportunity but haven't captured it.

But in my experience, the real number is even higher than 73%. Because when we actually map a finance function task by task - which we do for every client - we typically find that 80–85% of what the team does on a weekly basis is either fully automatable or can be done 10x faster with AI assistance.

The 15–25% that genuinely requires human judgment? Business decisions. Relationship management. Anything that requires interpreting ambiguous situations that don't fit a pattern. Strategic planning conversations. Those stay human. Everything else is fair game.

Why the automation hasn't happened

The honest answer is that it's not one thing. It's a combination of factors that compound each other:

1. The "good enough" trap

Finance teams are extraordinarily good at absorbing pain. If the close takes 14 days every month, the team just adjusts around it. The overtime becomes normal. The weekend board package prep becomes a ritual. Nobody measures the cost of this - partly because it's hard to quantify, and partly because the people suffering through it don't have the leverage to demand change.

We calculated the true cost for a client recently: a 45-person company whose finance team was spending approximately 340 hours per quarter on board reporting. At fully-loaded cost, that's around $68,000 per year. Just for one deliverable. They had no idea.

2. The IT dependency problem

In most companies, any automation initiative requires IT involvement. IT has a backlog. IT has priorities set by product, engineering, and sales. Finance waits. The wait becomes permanent.

This is one of the reasons AI-native tools have changed the equation so dramatically. Claude, used properly, can automate finance workflows without touching the IT backlog. The capabilities live in the finance function itself. A CFO with a Claude Pro subscription and the right system prompts can build automations that would have required a 6-month IT project two years ago.

3. Fear of accuracy

This one is understandable. Finance professionals are trained to be conservative and precise. The idea of letting a system generate numbers that go to the board feels risky. What if it's wrong?

But the framing is backwards. The question isn't "what if the automated output is wrong" - it's "how often is the manual process wrong, and what does it cost?" Manual processes have error rates too. They just feel safer because a human produced them.

The answer is hybrid: AI generates, human reviews. The CFO still signs off. The automation doesn't eliminate the human in the loop - it eliminates the 40 hours of mechanical work that precedes the 2 hours of human judgment.

4. Nobody owns the problem

Automation initiatives in finance tend to die in committee. The CFO sees the opportunity. The controller is skeptical. IT isn't prioritizing it. The CEO is focused on revenue. Three months later, the initiative has faded and the team is back to the 14-day close.

This is the organizational problem, and it's the hardest one to solve internally. It's also why most of our engagements start with a clear mandate from the CEO or CFO: we're doing this, here's the timeline, here's who owns it.

What the automatable 73% actually looks like

Let me be specific, because "73% of CFO functions" is abstract until you see it mapped out against your actual team's work.

When you add it up, a typical finance team at a $30M ARR company is carrying somewhere between 600 and 900 hours per year of work that could be automated or dramatically accelerated. At $100 fully-loaded cost per hour, that's $60,000–$90,000 in recoverable capacity - per year, every year.

The AI-first stack that's actually working

There's a specific combination of tools that, when configured correctly, captures most of the 73%. I can describe what our clients are actually running:

Claude for generation and reasoning. Everything that requires language - variance commentary, board narrative, policy documents, contract summaries, diligence memos - runs through Claude. At the Pro or Team tier, with well-built system prompts, the output quality is genuinely senior-level. Not a draft that needs heavy editing. An 80–90% final product.

Cloudflare Workers for automation pipelines. The recurring work - revenue forecast generation, close task routing, data pull from Stripe/Salesforce/NetSuite - runs on Workers with Cron triggers. It runs at 6am Monday morning, generates the forecast, posts to Slack. Nobody has to touch it.

Excel (properly instrumented) for the numbers. Excel isn't going away, and it shouldn't. But Excel connected to live data sources, with Claude generating the commentary, with automated distribution - that's a different tool than the spreadsheet someone refreshes by hand each week.

Your existing systems as data sources. You don't need to replace Salesforce, Stripe, NetSuite, or QuickBooks. You need to build the pipelines that pull from them automatically, at the right cadence, into the right format for Claude to process.

The 90-day path to capturing most of the opportunity

This isn't a 2-year ERP implementation. The automation gap in most finance functions can be closed in 90 days - methodically, function by function, with the right prioritization.

The sequence we use:

1. Map every finance task in week 1. Not the big-picture functions - every individual task, who does it, how long it takes, how often. Usually 60–90 discrete items. This is the foundation everything else is built on.
2. Score each task for automation potential using a simple rubric: Is the input data structured? Is the output format consistent? Does it require judgment or pattern recognition? High-scoring tasks go into phase 1.
3. Build the highest-ROI automations first. Typically board reporting, revenue forecast, and close task routing. These are high-visibility, high-frequency, and the ROI is immediately visible to leadership.
4. Instrument the remaining tasks over months 2 and 3 - expense processing, variance commentary, accruals. By month 3, the team's manual workload has typically dropped by 50–60%.
5. Redeploy the recovered capacity toward work that actually requires senior judgment: strategic analysis, investor relationships, business partnering with department heads.

What this means for your finance team

The uncomfortable part of this conversation is the staffing implication. If you can capture 73% of CFO function work through automation, what happens to the people doing that work?

The honest answer, from our experience: in most cases, companies don't shrink their finance teams - they stop growing them. The $20M ARR company that was about to hire a third accountant doesn't need to. The $50M company that was considering a VP of FP&A finds that their existing Director, properly equipped, can do the work.

The finance professionals themselves, in most cases, are relieved. The work that gets automated is the work they hate - the late nights refreshing spreadsheets, the manual reconciliations, the tedious board package assembly. What remains is the work they went into finance to do.

The companies that get this right will compound the advantage over time. Finance teams that have recovered 600 hours per year of capacity and redeployed it toward strategic analysis will make better capital allocation decisions, catch problems earlier, and close faster. That's a durable competitive advantage - not a one-time efficiency gain.

The 73% is sitting there. The only question is when you're going to capture it.