Case study
Argus — Multi-Agent Research Platform
A team of specialized agents that turns a research question into a sourced, contradiction-checked brief in under an hour.
01The problem
Northwind's analysts spent the first two days of every engagement doing the same thing: gathering market data, screening competitors, and assembling a landscape brief. The work was formulaic but couldn't be templated — every question needed judgment about what mattered and which sources to trust.
02What we built
We designed a multi-agent system: a planner decomposes the research question, worker agents run parallel searches and extract structured findings, a verifier cross-checks claims across sources and flags contradictions instead of averaging them away, and a writer assembles the brief with a full citation graph. A budget controller caps spend per question and reports where every token went.
03The outcome
A landscape brief that took two analyst-days now lands in 40 minutes with citations analysts can defend to clients. Northwind runs it on every engagement; analysts start at 'critique and extend' instead of 'blank page'.
