Workflow Playbooks
Complete engineering playbooks for the recurring workflows every product team should run: how to set them up with Claude Code at real scale, share the output with your team, keep them running, and what they cost.
Activation Funnel Diagnosis with Claude Code, Step by Step
When a step in your activation funnel drops, the diagnosis is a hypothesis-and-evidence loop, not a dashboard. Here is the full build: drop detection, a generator/verifier/critic agent trio, evidence wiring across analytics, tickets, and releases, and a registry that remembers every hypothesis ever ruled out.
Read the workflowApp Store and G2 Review Mining with Claude Code, Step by Step
Your users review every release in public, and the complaints cluster by app version if you let them. Here is the full build: ingestion per store, a structured extraction pass, theme clustering, a version by theme matrix, and a weekly report with a daily 1-star spike check.
Read the workflowChurn Analysis with Claude Code, Step by Step
Claude Code absolutely can run segment-level churn analysis on a 10,000-user product. Here is every step it takes: the data pipeline, the clustering, the subagents, the scheduled runs, and the publishing, written as a build guide you can follow today.
Read the workflowContinuous Competitor Monitoring with Claude Code, Step by Step
Track 15-25 competitors continuously, diff every watched page against its prior state, and ship a weekly brief with only the material changes. The full build: per-competitor state, crawl and distill, a materiality judge, parallel analysts, and scheduling.
Read the workflowConversation Intelligence for AI Products with Claude Code, Step by Step
Your agentic product generates thousands of conversations a week, and inside them are user sentiment, agent failure modes, and use cases you never designed for. Here is the full build: a cheap-model extraction pass, embeddings, continuous clustering with stable cluster identity, and a weekly report your team reads.
Read the workflowSynthesizing 50+ Customer Interviews with Claude Code, Step by Step
Fifty hour-long interview and sales call transcripts are roughly two million tokens, and the pile grows every week. Here is the full build: per-call extraction, a verbatim quote library your team can search, persona validation against real evidence, and a weekly synthesis that never goes stale.
Read the workflowProductBoard vs Claude Code for Feature Prioritization, Step by Step
Can Claude Code replace a dedicated prioritization tool? Here is the full build: intake extractors across four sources, embedding dedupe into a stable problem registry, RICE scores backed by real analytics queries, and a judge agent that rejects any score without evidence.
Read the workflowFrom Meeting Notes to Jira, Docs, and Decisions with Claude Code, Step by Step
Your notetaker already transcribes every meeting. What stays manual is everything after: actions filed with owners, decisions logged somewhere findable, customer follow-ups drafted, and last week's promises actually tracked. Here is the full build.
Read the workflowMulti-Audience Release Notes with Claude Code, Step by Step
Nobody reads release notes because everyone gets the same undifferentiated list. Here is the full build: one release feeding five audience-specific outputs, each written for what that audience does next, with a critic and a human gate before anything ships.
Read the workflowA Product FAQ Bot That Stays Current with Claude Code, Step by Step
PMs spend hours a week answering the same feature questions from sales and CS. Here is the full build: a citable knowledge repo, a Slack answering flow that refuses to guess, a gap loop that grows the KB, and a freshness pipeline that keeps it from going stale.
Read the workflowQBR and Exec Review Prep with Claude Code, Step by Step
Quarterly review prep is days of pulling numbers from five systems into slides. Here is the full build: a committed review template, source-wired pulls, a claims registry where every number carries its query, a judge agent that re-runs spot checks, and a deck draft that remembers what you promised last quarter.
Read the workflowBuilding a Sales Objection Library with Claude Code, Step by Step
Mine your own sales calls and your competitors' public reviews for the objections that actually kill deals, cluster them into a canonical set, and maintain evidence backed responses as a versioned, approved asset your reps trust. The full build, including the approval workflow that makes it trustworthy.
Read the workflowSelf-Serve Product Analytics with Claude Code, Step by Step
PMs wait days for data team turnaround or grind through SQL themselves, and chat-with-data demos fall apart the first time a number is wrong. Here is the full build: a committed semantic layer, a weekly metric pack, ad hoc answers that ship with the query attached, and a verifier agent that re-derives every number before anyone acts on it.
Read the workflowA Support Ticket Radar with Claude Code, Step by Step
Cluster today's support tickets, diff them against yesterday's baseline, and get a loud alert when a genuinely new problem appears. Here is the full daily build: export, extraction, stable clusters, alert thresholds, agents, and scheduling.
Read the workflowNPS Verbatim Analysis with Claude Code, Step by Step
Your NPS survey closes with 2,000 open text responses and the question that matters is not what themes exist, but which ones grew since last quarter. Here is the full build: a wave registry, a versioned theme taxonomy, score to theme joins, and a critic that refuses to headline a theme with n=12.
Read the workflowA Weekly Voice-of-Customer Rollup with Claude Code, Step by Step
Your customers say the same thing in tickets, sales calls, community Slack, and reviews, and nobody sees the whole picture. Here is the full build: five source extractors normalized into one record format, cross channel dedupe with embeddings, theme clustering with owners, and a weekly diff aware rollup.
Read the workflowAutomated Weekly Stakeholder Updates with Claude Code, Step by Step
Every Friday, a scheduled Claude Code run pulls the week from Jira or Linear, GitHub, and Slack, then drafts exec, team, and customer updates that a PM reviews in minutes. Here is the full build: source pulls, signal mapping, narrative memory, writer and critic agents, and scheduling.
Read the workflowWin/Loss Analysis at Scale with Claude Code, Step by Step
Your CRM's loss reasons are polite fiction, and the truth is sitting in call transcripts nobody rereads. Here is the full build: a per deal extraction pass that joins CRM notes with transcripts, dollar weighted theme clustering, and a rolling report PMs and sales ops actually use.
Read the workflow