v0.4  ·  customer intelligence

Stop summarizing customers.
Start modeling them.

Record real customer conversations. Watch them become structured, citable claims — the same way Anvil turns messy interviews into a queryable model.

live transform · v0.4 try it →
// raw_quote · customer interview 0 chars
modeling across 14 interviews
// claim_output · structured insight-0042
pain recurring high_urgency

Tools collapse at team ≥ 5 — top pain in this segment.

evidence
12 / 14
confidence
87%
sources
INT-0307 +9
segment
Mid-mkt PMs
Download for Mac
// how it works

From raw interviews to a queryable model, in three steps.

01 — Capture

Record live, or upload your existing interviews — and they get tagged as they come in.

Anvil ingests your interviews from Zoom, Granola, Otter, or paste-in notes. As they parse, pains, jobs, triggers, and objections get auto-tagged in your taxonomy. Nothing summarized away. Every quote keeps its timestamp and source ID.

14:02"Budget process eats my whole Tuesday."pain
14:08"I just want last quarter's number."jtbd
14:15"Tried Mode and Looker — both abandoned."
14:22"After our first analyst hire, things changed."trigger
14:28"We just bought another tool last quarter."obj
02 — Model

Patterns cluster, claims rank, and the model gets sharper with every interview.

Similar language clusters across the corpus. Recurring pains rank by frequency and confidence. Contradictions surface instead of getting flattened away. Add interview #15 and the whole model updates — no re-runs, no manual stitching.

Tools break at scale12/14
Plain-English Qs10/14
Per-seat pricing9/14
First analyst hire9/14
Workflow drag7/14
top claim
Tools collapse at team ≥ 5
12/14
contradicts
"We love per-seat" — INT-0314
1/14
03 — Query

Ask the synthetic segment anything — every answer cites the interviews behind it.

Test pricing, copy, positioning, objections. Anvil answers from your interviews, not the open internet. Click any claim to jump to the exact transcript moment, timestamp, and participant. No hallucinations. Insights your team can actually defend in a review.

Will per-seat pricing land here?
No. Per-seat is exactly the trigger this segment associates with "tools that punish team growth." Flat-tier or usage-based would land better.
⊙ INT-0307, 0312, 0319 · 87% conf.
backed by
INT-0307 · 14:02
"They all break the moment our team grows past five."
Mid-mkt PM · Series B
INT-0312 · 09:18
"It just falls apart once you bring in another analyst."
Sr PM · Series A
// faq

Questions, answered.

Notetakers summarize one interview at a time. Anvil models a corpus — clustering language across every interview, ranking pains by evidence, surfacing contradictions, and letting you query the segment as a whole. The output isn't a summary doc, it's a queryable model with citations.

No. Every claim Anvil produces is anchored to source quotes, timestamps, and participant IDs. If it can't cite the interview, it doesn't make the claim. You can click any insight and see the exact transcript moment behind it.

Live recording, uploaded transcripts (Zoom, Granola, Otter, manual), and pasted notes. Anvil parses speaker turns, timestamps, and metadata automatically. Bring your own taxonomy or use ours.

Product managers and researchers running real customer discovery — typically 10+ interviews per study. If you're already doing structured synthesis in Dovetail, Notion, or a giant spreadsheet, Anvil replaces that step.

Locally on your Mac by default. Anvil is a desktop-first product — your interviews don't leave your machine unless you explicitly sync them. Cloud sync is opt-in, encrypted, and per-workspace.

v0.4 is in early access for Mac. Click "Download for Mac" — we'll send you the build and a short onboarding. No credit card, no waitlist gatekeeping.

// ready

Customer knowledge should compound.
Not sit in a folder.

Built for product & research teams running real interviews.

Download for Mac