---
title: "Building Sales/GTM Workflows with Claude Code"
url: https://stacklist.com/card/6b78a2f9-9bbc-4577-aa04-70c9dad9b916
source_url: "https://x.com/chrispisarski/status/2082236016161677644?s=12"
stack: https://stacklist.com/c/business/stack/7c5a8832-3368-4e35-961e-f9273bcef8cc
summary: "Chris Pisarski shares a comprehensive guide on building sales and GTM workflows using Claude AI integrated with multiple tools like HubSpot, Slack, and Crustdata. The post covers ICP mapping, email infrastructure setup, signal-based list building, and lookalike audience creation with detailed implementation steps and prompts."
tags: "sales-automation, gtm-workflows, claude-ai, crm-integration, outbound-strategy, signal-based-targeting, email-infrastructure"
key_entities: "Chris Pisarski (person), Claude (technology), HubSpot MCP (technology), Slack MCP (technology), Google Workspace CLI (technology), Fathom MCP (technology), Crustdata MCP (technology), Instantly MCP (technology), ICP mapping (concept), GTM workflows (concept), signal-based targeting (concept), Proofpoint (technology), Mimecast (technology), Barracuda (technology)"
classification: "tutorial"
content_hash: "sha256:1cc5cd2c67b4e2370c6c58ae000a77d46e8c33ac65510016590022d614679232"
acp_version: "0.2"
token_counts_approximate: 1795
visibility: public
agent_accessible: true
status: "final"
---

# Building Sales/GTM Workflows with Claude Code

Chris Pisarski @chrispisarski this is how we built all of these sales/GTM workflows in-house with claude code step 1) connect claude to all of your tools, these are the ones we use internally: - Hubspot MCP (CRM) - Slack MCP (team comms + alerts) - Google Workspace CLI (docs, sheets, etc.) - Fathom MCP (call recordings) - Crustdata MCP (people and company data) - Instantly MCP (outbound) step 2) ask claude to build the workflows for you for example.. a) ICP / TAM mapping: - export all closed-won deals from hubspot: company, deal size, sales cycle length,... - if you don't have any closed-won yet: use best open opps + competitors' customers (scraped from their case-study pages / G2) - enrich all via Crustdata MCP: industry, headcount, headcount by department, geo, funding stage, tech stack, open roles,... - ask claude to run this prompt: "you have ./winners/ (one json per closed-won customer) and ./crm_export.csv (deal size + cycle length) 1. flag every attribute shared by 70%+ of winners, weighted by deal size 2. drop attributes any random b2b company would also match 3. write ./icp.md: hard filters (industry, headcount, geo, funding) + soft signals with weights (dept ratios, hiring, tech) + anti-icp (attributes of wins that churned or closed slow) 4. spawn a subagent to blind-score every winner against icp.md. 80% of winners must score 70+. loop until they do." b) mail infra + warmup: - never send cold from your main domain, buy 2-5 alternate domains to start (yourbrand-hq com, tryyourbrand com), 20-40 at scale - 2-3 inboxes per domain. person-first names (john@, not sales@), no numbers - dns on every domain: spf, dkim, dmarc + custom tracking domain (CNAME) - disable open tracking, pixel hurts deliverability, track replies only - warmup 2-4 weeks before sending anything. start 5-10/day per inbox and then start ramping up - warmup never stops: after ramp, ~20 cold + ~30 warmup per inbox daily, 40-50 total - capacity: 1,000 cold/day = ~50 inboxes across ~20 domains - before every campaign: verify the list (bounce rate must stay under 2-3%) + run an inbox placement test (primary vs promotions vs spam) - check all prospects that are behind secure email gateways (proofpoint, mimecast, barracuda). throttle or just delete those and set up a campaign on linkedin instead c) signal-based lists: 1. reverse-engineer which signals made your prospects reply - export your outbound history: every prospect ever contacted (instantly export + hubspot): date contacted, replied y/n, meeting y/n, won y/n + fathom as context - for each account, reconstruct what was true ON the day you contacted them via crustdata: headcount delta the quarter before, days since last funding, open roles matching your buyer titles at that time, new VP+ hire in the prior 90 days, posts made, etc. - ask claude to run this prompt: "you have ./outbound_history.csv (every account ever contacted: date, replied, meeting, won) 1. for each account, reconstruct the signal state at contact date via crustdata: headcount delta prior quarter, days since funding, open roles matching [titles], exec hires prior 90 days, posts, all relevant signals 2. calculate lift per signal: reply rate with signal vs baseline reply rate 3. calculate each signal's window: median days between signal and the replies it produced 4. read the reply threads per signal and extract the angle that worked 5. write ./signals.md: only signals with 1.5x+ lift, each with: lift, window, the proven angle. everything else gets deleted, not monitored 6. verify with a subagent: hold out 20% of history, check the ranking predicts reply rate on the holdout. loop until it does" - no outbound history yet: run the same analysis on closed-won instead (what signal states existed in the 90 days before each winner entered pipeline) - THEN set up a watcher via crustdata that watches for these signals in real-time, and add that company to your sequence in instantly d) lookalike lists off closed-won: - rank customers first: ACV × speed-to-close × expansion, minus churn (using connectors) - for each: reconstruct the company as it was when it bought, via crustdata (headcount, dept mix, funding stage, what they were hiring at that date) - pull the fathom transcripts from that deal: why they bought, in their words, and the proxy for that pain (e.g. "drowning in manual prospecting" → SDR headcount growing with no ops hire) - ask claude to run this prompt on every new closed-won: "1. check the customer ranks in the top third (ACV × close speed × expansion, minus churn) 2. reconstruct the company at purchase date via crustdata, not its current state 3. read the fathom transcripts from the deal. extract the buying trigger in their words + the observable proxy (what would this pain look like from the outside) 4. crustdata search: companies matching the at-purchase profile AND the pain proxy. exclude current customers, open pipeline, closed-lost < 6 months old 5. score 1-100 on firmographic match × pain-proxy match. keep 70+, cap at 25 6. verify with a subagent: blind-mix the 25 with 25 random companies that pass basic ICP filters. it must identify the real lookalikes 80%+ of the time. if it can't, the criteria are too generic, tighten and loop" - you can then push the list into a sheet (google workspace CLI), add it to hubspot, or send it to instantly e) champion tracking: - pull every contact on closed-won deals from hubspot: tagged champion/decision-maker + get more context from fathom - resolve their emails to linkedin profiles via crustdata (batch reverse-email lookup) - one-time backfill first: enrich all of them, diff current employer vs the deal's account - create a crustdata watcher on each profile, set up a slack channel to get a notification every time something happens: new funding round, new hire made by that company, web traffic, posts where they mention something specific - you can then set up claude workflows to react to every signal that is interesting for your industry ask claude: "pull fired watchers. for each job change: 1. enrich the new company + the new email of the champion using crustdata 2. score the new company against icp.md. below 70 → log it in hubspot, no alert 3. 70+ → create the company + contact in hubspot, tag champion-landed, assign the AE from the original deal 4. slack the AE: who they are, what they bought, the original deal size, what the new company does + a drafted first touch referencing the history using fathom as context" these are just some of the workflows you can power by connecting your stack to claude you can build any dream GTM/sales workflow yourself by just asking claude to go through all of these workflows and build something specific for your industry and company, with all the context it has about you Chris Pisarski @chrispisarski Jul 27 4 months later and the majority of YC founders i'm speaking with are still trying to hire the person that can do all of this with AI: - map out the entire ICP and TAM - build signal-based lists - build lookalike lists off closed-won accounts - track every champion who changes Show more