Public procurement
Tender monitoring and response
Catch relevant notices on BOAMP and TED, qualify them with frontier language models, and assemble the response pack.
- Weekly monitoring
- 6 h → 20 minWeekly monitoring
- Qualified opportunities
- 3xQualified opportunities
- Alert on publication
- Same dayAlert on publication
Context
A construction SME checked BOAMP by hand, twice a week. Good opportunities surfaced late, go/no-go was a gut call, and the technical proposal was always written against the clock.
Workflow
- 01
Collect
The open BOAMP and TED APIs are polled twice a day. Notices land normalised in one database.
- 02
Filter
Sorted by CPV code, region and estimated value. Noise is gone before anyone reads a line.
- 03
Qualify
A long-context frontier model reads the RC and CCTP end to end, pulls out the requirements, and scores the notice against past wins.
- 04
Go / No-Go
A summary hits Slack with the deadline, the required documents and the score. The call stays human.
- 05
Assemble
The standing document pack is assembled, certificates are checked for expiry, and the model drafts a first technical proposal grounded in the CCTP.
Stack
- BOAMP API
- TED
- n8n
- Claude Opus
- GPT-5
- Postgres
- Slack