For CFOs
Why does our AI give a different revenue number than finance?
Because the AI and finance are answering two different questions that share one word. Revenue before or after refunds. Booked or recognised. This month in which time zone. Nobody wrote the choice down once, so every tool made its own.
The usual fix is called a semantic layer: one shared dictionary of what each metric means, which every report and every AI assistant reads from. It is a good idea. It is also one more thing to build and run, and what decides whether it works is not the software. It is who is allowed to say what revenue means.
I keep seeing companies buy the dictionary before they agree on the words. Who signs off the definition of revenue where you work?
A working session on the data spend you cannot explain. It can be credited toward the Audit.
The hours spent reconciling figures are the first cost. About five minutes, no sign-up.
What to look at first
Put the answers side by side
Pick the five numbers your board sees: revenue, margin, churn, customer count, cost to win a customer. Ask finance, marketing and the AI assistant for each. Write the answers in one table. The gaps are the work.
One owner per number
One person who can say what it means and say no to a second version. Often in finance, sometimes not. Without a name, every fix lasts until the next dashboard gets built.
Price the dictionary before you buy it
A semantic layer needs people to build it and keep it current, every year. Ask what it costs to run, which reports it replaces, and which of the three answers it ends. If it replaces nothing, it adds cost.
The longer version: One number, one meaning
Not your question?
- CFOHow do I know if our data and AI spend is paying off?
- CEOHow do you build a data strategy that actually gets finished?
- CFOWhat should we ask before we build an AI project?
- CFOWhat does a dashboard, pipeline or AI model cost to keep running?
- CFODid our data cost cut actually save money?
- CFOWhat does data capital allocation mean for a CFO?
- CFOI have just inherited the data and AI budget. Where do I start?
- CFOWe announced an AI budget. What return should we hold it to?
- CFOWhat is our AI spend returning, and what should we stop paying for?
- CFOHow much of revenue should we spend on data?
- CFOShould we stop this AI pilot or keep going?
- CEOHow should a CEO judge the head of data?
- CFOShould we approve this AI purchase?
- CEOWhat should a CEO ask about data and AI?
- CEOWhy has our data spend changed so little?
- Data teamHow do I show leadership where my data team's time really goes?
- Data teamHow should leadership measure my data team?
Prefer something to print? One-page sheets: reviewing an AI business case, seven questions for every budget line, hours lost to data workarounds.