Have you ever seen an AI presented on the big screen in a meeting? Everyone claps. Six months later, nobody uses it.
That is the AI that impresses. There is another one, less pretty and far more useful: the one that pays the bills.
What it looks like
It lives inside a process that already exists and moves a number that finance tracks.
In the projects I led, that meant an average ROI of 11.6x. There was no magic. There was focus on the right number.
How do you find one?
I use three questions. Where do we lose time repeating the same thing? Where do we decide in the dark, with no data in front of us? And where does the right information take too long to reach the right person?
Notice that none of them starts with "which model do I use". They all start with "where does it hurt".
The market says the same
MIT measured that about 95% of AI pilots don't move results. And IBM points out that the return rarely comes from the model. It comes from embedding AI in a process that finance tracks.
AI that pays the bills becomes a line in a spreadsheet.
For me, good AI is the one that hits the right pain and disappears into the process. The number goes up and nobody notices the technology.
And in your company, which AI is running: the big-screen one or the spreadsheet one?
