Velq
An AI agent system that turns a scattered sales process into revenue.
Context
SMEs know AI could help them sell more, but they don’t know where to start: no time, no internal expertise, no execution plan. I wanted to build the concrete answer rather than one more theory.
The problem
A scattered sales process — forgotten follow-ups, poorly qualified leads, reporting done by hand. Generic AI solves nothing until it is wired into the actual sales motion.
What I did
I designed Velq, a productised AI service for SMEs and sales teams, deployed on Cloudflare Workers and Supabase. The core is a multi-agent system: an orchestrator breaks down the work and routes it to specialised agents — company research before a call, lead qualification against an ICP, sequence strategy, meeting prep. I enforced a source-honesty rule: every agent output is tagged [verified] or [hypothesis], so a guess is never presented as a fact.
Results
A system that runs, not a mockup: documented and versioned agents, published B2B content, premium design, trilingual FR/EN/DE interface.
What I took away
Orchestrating agents is product management applied to AI: deciding who does what, where to place a guardrail, what still deserves a human. That is exactly the job I am aiming for.
Highlights
- 17 specialised agents driven by a central orchestrator
- Anti-bullshit guardrail: every claim tagged [verified] or [hypothesis]
- Live in production, trilingual FR/EN/DE interface
What I took awayA useful AI agent is not another chatbot — it is a delegated decision, with a guardrail.