DELOS Analytica · Streaver

The problem

Public affairs and communications teams struggle to track thousands of stakeholders, regulators, journalists, NGOs and policy threads, and translate that noise into board-ready intelligence.

What was deployed

Vendor

Streaver

Products

OpenAI, LexisNexis, Opoint, OpenSearch

Technique

LLM

Build or buy

Built in-house or to order

Deployment

Custom build

Data used

News from LexisNexis and Opoint, Swiss Parliament's open data (221,878 affairs, 57,574 labelled events).

Scale

~25 paying enterprise customers, 221,878 parliamentary affairs indexed, ~25 languages

Timeline

4 months

What changed

Each row is quoted from the source. Figures we could not find on the page in those words are marked — they are kept, not deleted, so you can judge them.

Difficulties and limits

• MVP had to be live in sixteen weeks, solving four problems simultaneously.
• AI matching needed to earn trust, avoiding hallucinations.
• Ingestion had to normalize data from multiple providers into one model, resilient to schema drift, without a six-month data engineering build-out.
• Interface needed to be intuitive for non-power users (heads of public affairs).
• Required Swiss-grade governance (multi-tenant, roles, row-level filtering, SSO, audit, GDPR) from day one.

• A custom crawler built in early 2025 (Firecrawl-based) was retired in September 2025 as it never earned its keep next to commercial providers.
• The monitoring pipeline was rebuilt twice (three major versions in total), with the first two versions being 'expensive drafts'.
• An architecture decision record for a third-party extraction vendor was never implemented.
• The evaluation harness for AI pipelines landed in August 2026, two years late, meaning quality was initially managed by score thresholds and human triage rather than measurement.

Sources

• streaver.com — vendor self-report, 2024-12-01