The tech stack of the future for strategy and finance teams.

Storage. Production data in a governed warehouse. This layer is largely solved. The open question is access, not location.

Context. A semantic layer between the warehouse and everything upstream. Defined metrics. Approved dimensions. Row-level security. Business language on top of raw tables. Agents need this to query safely and get the same answer twice.

LLM. Where the daily workflow moves. You describe the question. The model pulls data, interprets it, drafts narrative, maybe renders a first pass. Human judgment still matters. The starting point is a conversation, not a blank grid.

Charting. Turn AI output into live, editable charts that travel across surfaces. Power-user control. Data attached. Fit for recurring reporting. Chartbuddy is what we are building for this layer.

Reporting. Distribution stays fragmented. Board decks. Dashboards for monitoring. Written memos. Agent-generated briefs you review and send. The stack has to support multiple endpoints from the same analysis.

Not spreadsheets. Not decks. The glue work. Rebuilding the same monthly pack from scratch. Screenshotting a dashboard tile and reformatting it. Waiting on a pull an agent could run in minutes. Pasting chart images with no data behind them.

What stays is the human part. Judgment. Narrative. Knowing which number matters this quarter.

Data in the warehouse. Metrics defined in the semantic layer. An agent pulls a revenue bridge and flags anomalies. Charts get finalized and land in the board pack. A trimmed version feeds a dashboard for monthly monitoring. Same numbers, three surfaces. That loop is the workflow, not the exception.

Timing and winners, mostly. Semantic layers will consolidate. Some teams will skip dashboards for ad-hoc work once agents are good enough. Memos might eat more of the board pack than decks. Maybe. Document-first reporting would not be ideal if you sell charting software. You win some, you lose some.

What seems harder to argue with: agents need a governed context layer above the warehouse. And if your deliverable is still a chart-heavy deck, a static image from chat is not a long-term answer.

Not "which AI tool should we buy?" but "where does insight still die between the warehouse and the stakeholder?" For most teams I talk to, the answer has not moved much. The data is fine. The last mile of the report is where time goes.

If you are mapping the same transition today, this piece on fragmented FP&A workflows is a useful companion read.

• The FP&A Reporting Workflow Is Fragmenting
• How to make charts in Claude (and where it still falls short)
• Stop Rebuilding Your Monthly Report From Scratch