60% of a research workflow, today
An honest capability inventory. What weev can already do end-to-end, what's a few weeks away, and what's deliberately not on the roadmap.
We get asked a lot: "so what does weev actually do today, and what's marketing-speak?" Fair question. Here's the honest scorecard.
What's live in production
Conversational AI employee. Brain service orchestrates every other capability. You talk to her, she routes the work, she asks for confirmation before launching anything expensive or irreversible.
Voice interviewing. weev joins Zoom and Teams meetings under a configurable bot name. Real-time speech-to-speech via OpenAI Realtime. Dynamic probing rules. Turn-by-turn transcripts auto-indexed into the knowledge base.
Qualitative analysis. Five-stage pipeline: questionnaire parsing → response cleaning → per-question theme extraction → K-means clustering → persona generation. Personas come with named clusters, traits, pain points, and representative quotes.
Quantitative analysis. Seven-stage pipeline: data context → descriptives → cross-tabs → significance testing (with Bonferroni and effect sizes) → segmentation → narrative insight → report assembly with chart generation.
Reporting. PowerPoint decks generated from any analysis payload, with brand tokens applied per client. Reveal.js interactive HTML decks for the modern audiences.
Knowledge management. Every past study indexed in pgvector. Semantic search across questionnaires, analyses, personas, and transcripts. Every chat with weev is grounded in the org's prior work.
What's coming next quarter
- Questionnaire designer (today she parses, soon she writes).
- Online survey hosting + live field dashboard.
- Panel provider integration (Prolific first).
- Fraud / quality layer.
- Key driver analysis.
What we're deliberately not building
Specialist methods with low pull (eye-tracking, biometrics, structural equation modelling). We'll build them when a paying customer drags them in. Speculative breadth is a great way to ship a mediocre everything.