Situation
A European industrial leader needed structured competitive intelligence across a defined set of competitors and a defined set of signal axes - news, content, hiring, pricing, patents, market movements. Existing commercial solutions were expensive and delivered undifferentiated coverage. Meanwhile, most of the underlying signal was already publicly available.
Friction
Two frictions had to be resolved before any agent could be deployed. First, the source policy: which public signals could be collected, under which terms of service, and under Swiss and EU data-protection rules. Second, the governance policy: what the agent was allowed to do autonomously, and what had to be human-reviewed - especially for anything related to people movements.
- Sourcing had to be defensible under GDPR / nFADP.
- Terms of service had to be respected by design.
- Anything touching individuals required explicit human review.
Intervention
TSG built an API-first collection layer restricted to authorized public sources, with robots.txt respected by construction and a full audit journal. The agent's autonomous scope was limited to structuring and classifying signals; any output touching individuals was routed to human review before it could enter internal briefings. A named business sponsor owned the register of processing under Swiss and EU rules.
- API-first collection with source whitelist.
- Robots.txt and terms of service enforced by design.
- Autonomous scope limited to non-personal signal.
- Human review mandatory for anything touching individuals.
- Register of processing owned by a named sponsor.
Operating model
The agent runs on a fixed cadence and produces a structured brief per competitor and per axis. Every entry is traceable to its source. The business owner approves the addition of any new source. Any change to the autonomous scope requires an explicit governance decision. The system is designed to be reviewable end-to-end by an internal or external auditor.
Evidence
The evidence tracked is qualitative. The competitive intelligence output is now defensibly produced, source-traceable, and reviewed on a fixed governance cadence. The economics moved from a recurring commercial subscription to a controllable internal capability. Specific commercial figures are intentionally not disclosed in this brief.
Cas d’usage et impact
Expensive, undifferentiated commercial competitive intelligence, difficult to defend under Swiss and EU rules.
Solution: Governed internal agent, API-first, source-whitelisted, with human review on any output touching individuals.
- Source traceability: Partial
- Governance review cadence: Ad hoc
- Autonomous scope on personal data: Undefined
- Source traceability: Full, per entry
- Governance review cadence: Fixed
- Autonomous scope on personal data: Explicitly excluded
FAQ
Why not just use a commercial tool?
Because most of the underlying signal is public, and because a governed internal capability is easier to defend under Swiss and EU rules than a black-box commercial pipeline over which the client has no runtime control.
How is people-movement intelligence handled?
It is not handled autonomously. Any output touching individuals is routed to human review under an explicit register of processing. This is a design decision, not a workaround.
Is this replicable across industries?
The architecture is. The source list, the governance owner and the sensitive axes have to be redefined per client - which is where the real work is.
Conclusion
Competitive intelligence does not need to be an opaque, recurring cost. Built with the right sourcing discipline and the right governance boundaries, it becomes a controllable internal capability - defensible under Swiss and EU rules, and directly aligned with how the executive committee wants to see the market.