Product discovery
Archeus clusters your interviews, tickets, and reviews into root-pain themes with quotes and ARR attached — then drafts the PRD, the ADR, and the acceptance criteria that survive engineering.
Q3 Feedback — 212 signals
Interviews · Tickets · Reviews
Root-pain clusters
Onboarding dead-ends
CSV export limits
Mobile parity
Acceptance criteria ready
14 scenarios · 9 edge cases
Real workspace preview — clusters, the decision record, and the handoff stay traceable.
Built for teams that review before they ship
In action
A real working view — reviewable before anything ships.
212 raw signals become six themes — each one keeps the customer's words and the ARR it touches.
How it works
A practical, reviewable workflow—not a black box.
Interview notes, tickets, and reviews are gathered around one question — with account and ARR metadata on every line.
Prompt
“Collect Q3 feedback from interviews, tickets, and reviews around one question: where do users get stuck?”
Pipeline
Pull sources
Interviews · tickets · reviews
Normalize snippets
212 signals kept
Attach metadata
Account + ARR on each
Verifiable output
Root-pain themes emerge with the customer's words intact — ranked by mentions, affected accounts, and the revenue they touch.
Prompt
“Cluster the 212 signals into root-pain themes — keep the customer's words.”
Pipeline
Group by root pain
6 themes, not keywords
Rank by ARR at stake
$2.05M total
Bind quote evidence
Every theme cites sources
Verifiable output
Trade-offs are scored against your goals and the decision is funded — the ADR keeps who chose, why, and what would revisit it.
Prompt
“Frame the trade-off: which cluster do we fund this quarter, and why?”
Pipeline
Score vs activation goal
Export cluster wins
Record decision + risks
ADR-014 drafted
Name what would revisit it
Volume > 5M/day
Verifiable output
Evidence separates from assumptions
Trade-offs recorded in ADR-014
Mobile parity deferred — revisit Q1
The PRD opens with a cited problem, keeps scope honest, and ends with metrics engineering can instrument before the build.
Prompt
“Draft the PRD with the problem cited, scope bounded, and metrics defined.”
Pipeline
Cite the problem
INT-142 · 33 quotes
Bound the scope
2 in · 1 explicitly out
Define metrics
P50 < 8s · tickets → 0
Verifiable output
PRD — Export v2
Accounts can’t complete the monthly close in-product — the 50k cap forces chunking.
Stakeholder summaries and Gherkin scenarios hand off clean — the PM's claims survive contact with the backlog.
Prompt
“Turn the work into a stakeholder summary the team can act on Monday.”
Pipeline
Summarize evidence → decision
One page, sources linked
Attach acceptance criteria
Gherkin, sprint-ready
Distribute for review
PM owns the send
Verifiable output
Capabilities
Clusters keep the customer's own words with account and ARR attached — roadmap debates start from shared evidence.
every theme links to its source quotes — id CL-2026-Q3
Each funded decision carries its trade-offs, its owner, and the condition that would revisit it — in a durable ADR.
ADR-014 — async export queue
Revisit condition: volume > 5M/day
Acceptance criteria arrive as Gherkin scenarios mapped 1:1 to scope — the PRD's claims survive contact with the backlog.
Engineering reviews criteria, not paragraphs — and the PM’s summary links back to every source.
Handoff — export_queue.feature
Human checkGiven
2M-row account requesting CSV
When
export is queued within 1s
Then
file + notify inside 5 min
Where it helps
Start with the workflow you already repeat — the pipeline maps to it, not the other way around.
A quarter of interviews becomes an evidence map — themes, quotes, and ARR, ready for prioritization.
Assumptions, constraints, and revisit conditions become visible before the bet is sized.
From top cluster to cited PRD and sprint-ready criteria — in an afternoon, not a fortnight.
Product, marketing, and CS read the same evidence-linked summary — no translation losses.
Why teams switch
Roadmap debates restart from memory and the loudest anecdote.
Themes arrive with quotes, accounts, and ARR — the debate starts from evidence.
The PRD's problem statement is a paraphrase nobody can verify.
Every problem statement cites its interview or ticket — 33 quotes behind this one.
Architecture decisions live in a Slack thread nobody can find.
ADRs record the trade-offs, the owner, and the condition that would revisit them.
Acceptance criteria get invented during the sprint.
Gherkin scenarios map 1:1 to scope — engineering reviews criteria, not paragraphs.
Enterprise Standards
Customer prompts, uploaded business records, connected databases, and generated outputs are strictly ephemeral and never used to train base or commercial foundation models.
Automations, code evaluation, and browser executions occur in isolated, scoped environments with strict egress policies and encrypted secrets management.
Consequential actions—such as customer communications, database writes, or live deployments—require deliberate human confirmation with comprehensive audit logs.
Questions
Start with one real workflow, keep people in control, and build from what works.