The First 3 Workflows to Build in Claude Code

If you have Claude Code installed and your first .md file built, you are ready to start building.

Now the question is what to actually build with it.

I am going to give you three workflows in order. Each one builds on the last. I am including the actual terminal commands and prompts so you can run these today.

Workflow 1: CRM Data Cleanup

Start here. Your data is dirty and everything you build on top of dirty data will underperform. I have watched operators build beautiful enrichment pipelines on data that was 30% duplicates and 20% wrong titles. The pipeline ran perfectly. The output was garbage.

Open your terminal. Navigate into your Claude Code project folder where your icp-context.md file lives.

cd ~/claude-projects

Export your CRM contacts as a CSV and drop it into this folder. Then launch Claude Code.

claude

Now type this prompt:

Look at contacts.csv. Deduplicate rows by email address.
Normalize all job titles so variations like "VP Sales",
"Vice President of Sales", and "VP, Sales & Marketing"
map to standardized categories. Flag any row where the
email domain is a personal account like gmail, yahoo, or
hotmail. Remove any contact that does not match the ICP
criteria in icp-context.md. Save the cleaned version as
contacts-cleaned.csv and give me a summary of what was
removed and why.

Claude Code reads the CSV, reads your icp-context.md file, writes a script, runs it, and gives you the cleaned output. What used to take a VA an afternoon runs in under a minute.

Check the summary. If contacts were incorrectly removed, refine your icp-context.md file and run it again. That is the compounding. Every time you clean the criteria, the next run gets sharper.

Workflow 2: ICP Enrichment and Scoring

Now that your data is clean you can enrich and score it. This is where Plan Mode matters.

Launch Claude Code with the Plan flag so Opus handles the architecture and Sonnet handles the code generation

claude --plan

If you have call transcripts from Gong or Fathom, drop them into your project folder as text files or PDFs. Then type this prompt:

I want to build an ICP enrichment and scoring pipeline.

Read icp-context.md for my qualification criteria.
Read data-sources.md for my API configuration.

Step 1: Analyze the call transcripts in /transcripts
to identify objection patterns, win/loss signals, and
the language buyers use when they describe their pain.
Output a scoring criteria document based on real
patterns from these calls.

Step 2: Take contacts-cleaned.csv from the last
workflow. Enrich each contact using the Apollo API
(direct API call, not MCP — I need token efficiency
for bulk processing). Pull firmographic data, tech
stack signals, funding status, and headcount.

Step 3: Score each contact against the criteria from
Step 1. Bucket into Tier 1 (high fit + signal),
Tier 2 (good fit, lower signal), and Tier 3 (nurture
or drop).

Step 4: Filter out existing customers using Stripe MCP.
Filter out already-contacted leads using Gmail MCP.

Step 5: Save the output as scored-leads.csv with
columns for score, tier, enrichment data, and the
reasoning for each score.

Use auto-mode but pause before the final save so I can
review the scoring output.

That prompt chain references your .md context files, uses direct API for bulk enrichment, routes filtering through MCP where it makes sense, and pauses at a checkpoint before saving. That is the architecture.

If it breaks midway through because of a rate limit or API timeout, Claude Code's session resume picks up exactly where it left off. You do not lose progress.

Review the scored-leads.csv. If the scoring does not match your intuition about which leads are actually good, update your icp-context.md file with more precise criteria and run it again. The .md file gets smarter. The pipeline gets sharper. That is the loop.

Workflow 3: Automated Sequence Generation

Now you have scored, tiered, enriched leads. Time to generate messaging.

Make sure your messaging-logic.md file is in the project folder. This is the file that maps your angles per ICP segment, your proof points, your objection handling, and your CTA logic by sequence position.

Launch Claude Code and type:

Read scored-leads.csv, icp-context.md, and
messaging-logic.md.

For each Tier 1 lead, generate a 3-email cold
sequence. Rules:

- Each email under 70 words
- Soft question-based CTA (no hard asks in email 1)
- Lead with the messaging angle mapped to their
  ICP segment in messaging-logic.md
- Include a concrete front-end offer (audit, playbook,
  calculator, or teardown) based on their pain points
- Generate 3 variations per email: simple, niche-aware,
  and hyper-specific
- Include Clay-compatible merge variables for
  {{first_name}}, {{company}}, {{pain_point}},
  {{tech_stack}}
- Generate spintax subject lines (3 variations each)

For Tier 2 leads, generate a 2-email templated
sequence using the same rules but less personalization.

Drop Tier 3 entirely.

Output everything as sequences-ready.csv formatted
for import into Instantly.

Use auto-mode but pause before final export so I can
review the sequences.

Claude Code is not generating cold emails from a generic prompt. It is reading your scored lead data, your ICP criteria, and your messaging logic together. It knows which segment each lead falls into, which angle to lead with, which proof points support that angle, and what CTA makes sense at each sequence position.

Review at the checkpoint. If the messaging sounds off for a specific segment, update your messaging-logic.md file. Run again. The output improves because the logic improved.

It all starts with the right sequence.

Clean your data first. Then enrich and score it. Only then generate messaging. Each step feeds the next, and your .md files get tighter every time you run it.

Don’t try to build everything in one go. Get the first workflow working, then layer the next pieces on top over the next few days. You’ll feel the difference quickly.

Once it’s connected end to end, things start compounding.

That’s where the real leverage is.

Stay safe out there.

Much Love,

Benjamin Reed

Revyops & NextGen Founder

P.S. Check out ReyvOps at Revyops.com.

P.P.S. Want to help me on my journey to build RevyOps into a $1 Million SaaS? Please share this newsletter with your network so they can follow along. Anyone can sign up 100% free using this link: Click here

The Builder's Playbook is a series by Benjamin Aaron Reed on the tools, systems, and workflows that give founders and GTM operators an unfair advantage. New issues drop regularly.