---
title: "OSINT for CFOs: reading the market before it reaches the order book"
url: "https://obserbi.com/blog/osint-for-cfos"
description: "Your forecast is built from internal history. The disruptions that break it are visible outside the company first — in public sources, weeks earlier."
---

# OSINT for CFOs: reading the market before it reaches the order book

July 15, 2026·3 min read·ObserBI

Every financial forecast is an extrapolation of internal history, and every serious forecast error comes from outside it. The supplier that files for restructuring, the customer whose own market collapses, the port that stops moving containers, the regulation that changes a product’s landed cost — none of these appear in your ERP until the moment they become an invoice. They are, however, usually visible in public sources weeks earlier.

## Why finance stopped reading the outside world

Not because CFOs stopped caring. Because the work does not scale. Watching two hundred customers, eighty suppliers and four regulatory regimes is a full-time reading job that produces nothing measurable most weeks, and it is always the first thing dropped when the close is late.

So it gets replaced by a proxy: the sales pipeline, which is the outside world filtered through your own sales team’s optimism and lagged by a quarter.

## What machine-assisted OSINT changes

Agents are good at exactly the part humans find intolerable: reading a lot of low-value text continuously and noticing the rare item that matters. For a finance function the useful sources are unglamorous and almost entirely public:

-   **Company registers and insolvency notices** for customers and suppliers.
-   **Procurement portals and tender results**, which show where your customers’ revenue will come from next year.
-   **Regulatory and standards publications** that change product requirements or duty rates.
-   **Logistics and commodity data** — freight rates, port throughput, energy and input prices.
-   **Job postings**, an underrated leading indicator: a customer that stops hiring in a division is telling you about next year’s volume.

None of this is exotic intelligence work. It is reading, at a volume that only became affordable once a machine could do the first pass.

## From signal to number

Collection is the easy half. The discipline is turning a signal into something a forecast can consume, and that means three unglamorous steps.

**Attach it to an entity.** A news item about a company only matters if the system knows that company is your third-largest customer, or supplies the component in your highest-margin product. Without an entity graph you have a news feed, not intelligence.

**Quantify the exposure.** “Supplier X is in trouble” is not actionable. “Supplier X provides the single-sourced component in 18% of Q4 revenue, and switching qualification takes eleven weeks” is a decision.

**Model the scenario, not the headline.** The output that belongs in a forecast is a probability-weighted range with the trigger named — so that when the trigger fires, the finance team already knows which branch of the plan they are on.

## Keeping it honest

Two failure modes make OSINT worse than useless in a finance context.

The first is **false confidence**: a model that reads a rumour and reports it with the same tone it uses for a filed insolvency notice. Every conclusion has to arrive with its sources and its confidence, or it will eventually poison a board discussion.

The second is **scope creep into people**. Monitoring companies is ordinary commercial diligence. Monitoring named individuals is a different activity with a different legal footing under GDPR, and a finance system has no business drifting into it. Draw the line at the legal entity and keep it there.

## The point

The internal loop tells you that margin is drifting. The external loop tells you why, and how long you have. A CFO agent that only reads the general ledger will always be explaining last quarter with great precision — which is exactly the failure mode we set out to fix.

-   OSINT
-   forecasting
-   risk
-   supply chain

## Keep reading

-   [The autonomous CFO is not a dashboard](https://obserbi.com/blog/autonomous-cfo-is-not-a-dashboard)
-   [What the EU AI Act actually asks of a finance model](https://obserbi.com/blog/eu-ai-act-finance-models)