Capability case study · Demand intelligence
In the environmental project economy (solar siting, stormwater, remediation, compliance), the work is triggered by public events long before it's contracted: permits filed, interconnection-queue movements, RFPs issued, environmental notices published, grant awards. The signal that a project exists is out there, in the public record, weeks or months before it becomes a competitive bid. We built a system that finds those signals and turns them into pipeline.
The problem
Those signals are scattered across thousands of sources, and the firms best positioned to win the work either find out late or never find out at all. Generic sales-intelligence tools index companies and people, not projects and regulatory events. And monitoring the sources by hand is a full-time job no small engineering firm has the capacity to staff.
So the demand is visible in principle and invisible in practice. The project is on the public record; the firm that should be first to it is still waiting for it to surface as a bid, by which point the positioning is already lost.
What we built
A demand-intelligence system that monitors the public record continuously, identifies real project signals, and turns them into pipeline-ready opportunities delivered straight into the client's CRM.
It runs unattended. New project signals surface, get classified and de-duplicated against what's already known, and arrive as structured records the sales team can act on, each one carrying its provenance, so it can be trusted and followed up.
How it works, in plain terms
The system deliberately separates three things that most scrapers blur together: the raw evidence, the interpretation of it, and the real-world project it points to.
1. Monitor. It polls a register of public sources on a schedule and ingests raw documents, stored immutably with their source, fetch time, and original text, so every downstream claim is traceable back to its evidence.
2. Classify. An AI pass reads each document and decides whether it contains a real project signal, and of what type, filtering the noise of the public record down to the events that matter.
3. Extract. A second AI pass pulls the structured detail (project, location, stage, estimated scale, key dates) into a consistent schema.
4. Resolve and de-duplicate. Signals are matched to the real-world projects they describe, so the same project surfacing across five sources becomes one clean record, not five noisy ones.
5. Deliver. Qualifying projects land in the client's CRM tagged with a dedicated lead source, where they feed a project-aware outreach sequence built specifically for them.
Every signal carries three timestamps (when the event happened, when it was published, and when the system fetched it), which makes speed-to-signal a measurable property of the system rather than a guess.
What makes it different
Projects, not companies. It tracks demand events, not org charts: the pre-bid signal window where positioning is actually won.
Built for the underserved buyer. It's designed for environmental firms too small for enterprise market-intelligence platforms and too busy to monitor the sources themselves.
Trustworthy by construction. Immutable source records, confidence-scored classification, and human review for low-confidence items keep the feed clean. A duplicate or misattributed project erodes trust faster than a missing one, so the system is built to earn the sales team's trust on every record.
Operationally integrated. It doesn't stop at a list. Signals arrive inside the CRM the firm already sells from, ready to work.
The stack
- Python pipeline
- a two-model AI pass (lightweight classification, then deeper extraction)
- an immutable source store
- scheduled cron worker on managed cloud hosting
- CRM delivery integration
What the firm gets
Instead of finding out about projects when they hit the bid stage, the firm sees them as they enter the public record: classified, de-duplicated, and dropped into the CRM with their provenance attached, ready for the sales team to work. The advantage isn't a longer list. It's being early, on the right projects, with the evidence to act on them.