Look at how the traditional finance intelligence stack is assembled, layer by layer.
At the bottom is the system of record — the ERP — which stores what happened. On top of that goes a transformation layer; ETL pipelines into a data warehouse which cleans and reshapes the data on a schedule. On top of that goes a visualization layer with BI tools and dashboards which renders the reshaped data into charts. And then, at the very top, sits you: expected to know your question in advance, request the report, and wait for the next refresh to answer it.
Every layer in that stack is optimized for the same thing: reporting what already happened, on a cadence. It is a magnificent machine for producing last month’s numbers on a predictable schedule. It is almost useless for answering a question you didn’t know you’d have until thirty seconds ago in a board meeting.
That’s the backward part. The architecture assumes the scarce resource is rendering ie. turning stored data into a chart. But in 2026 the scarce resource isn’t rendering. It’s reasoning meaning specifically connecting what’s happening across systems, in the moment, and knowing what to do about it. We built the whole stack to optimize the thing that stopped being the bottleneck years ago.
Read the full article on Substack