The 6 Workflows That Replaced a Team

Pull campaign analytics for all campaigns that ran in [MONTH] [YEAR] under the [WORKSPACE NAME] workspace.

Return:
- Total emails sent
- Reply rate
- Number of interested replies
- Interested rate
- Top performing campaign with: vertical/angle, send date, emails sent, interested replies
- Runners-up (next 3 verticals) with the same metrics

Emails sent: ~100,000
Reply rate: 2.15%
Interested replies: 113
Interested rate: ~10%

Top performing campaign: influencer marketing agencies
Send date: August 12
Emails sent: 12,000
Interested replies: 19+

Runners-up: podcasting, accounting, SaaS general

Use Case 1: Reporting & Analytics

The problem. you're jumping between your sequencer, your CRM, and a spreadsheet trying to figure out what's actually working, copying numbers and formatting them so you can compare angles. for most agency founders the data ends up sitting there. you have a rough sense of what's happening, but you never get clean, real-time numbers to operate off of. at scale, that reporting gap costs you. you either over-invest in segments that aren't working or kill campaigns that would have converted with more time.

The Claude Code prompt:

Pull campaign analytics for all campaigns that ran in [MONTH] [YEAR] under the [WORKSPACE NAME] workspace.

Report:
- Total emails sent across the window
- Aggregate reply rate
- Total interested replies
- Aggregate interested rate
- Top performing campaign by interested replies, including: vertical/angle, send date, emails sent, interested count
- Top 3 runners-up by vertical with the same metrics

Then explain why the top campaign likely outperformed (offer, list, angle) so I know what to re-run.

How to run it. load your sequencer API key (we use EmailBison) and drop the prompt into Claude Code, then swap the bracketed window and workspace. expected output is a top-level snapshot (sent, reply rate, interested), the winning vertical, and runners-up. ask follow-ups to go more detailed: "why did influencer marketing agencies outperform SaaS general? what was different about the offer?"

Edit slots. [MONTH] [YEAR] is your reporting window. [WORKSPACE NAME] is your sequencer workspace.

What "good" looks like. our actual run for August 2025 under one client workspace returned roughly 100,000 emails sent, a 2.15% reply rate, 113 interested replies, and a 10% interested rate. the top campaign was influencer marketing agencies on August 12, with 12,000 sent and 19+ interested. runners-up that month were podcasting, accounting, and SaaS general.

the report tells you what to re-run for the next 2 weeks. before Claude Code, that query took hours of manual digging through report pages.

Use Case 2: Loading Campaigns

The problem. loading a campaign manually is dead time. you've got the lead list, the copy, the sequences, and the settings ready to go. then you spend the next hour going line by line, uploading contacts, mapping fields, and double-checking everything landed in the right place. none of it requires real thinking. every hour spent on this is an hour not spent growing the business.

The Claude Code prompt:

Load a new campaign into [SEQUENCER NAME] using the following inputs:

- Campaign name: [CAMPAIGN NAME]
- Lead list: [PATH TO CSV OR LEAD LIST REFERENCE]
- Sender emails: [USE DEFAULT SENDER ACCOUNTS]
- Sequence copy:
  - Subject line: [SUBJECT LINE]
  - Email 1: [EMAIL 1 BODY]
  - Email 2: [EMAIL 2 BODY]
- Schedule: [USE DEFAULT SCHEDULE]

Keep all other campaign settings the same as my standard defaults.
Confirm campaign creation and return the campaign ID once loaded.

How to run it. connect your sequencer API and give Claude Code one CSV plus the email copy. it loads the campaign using your standard settings (same sender accounts, same schedule, same defaults). you stop finicking with form fields, and the entire load process collapses into: push in the list, push in the copy, give the green light to launch.

Edit slots. [SEQUENCER NAME] is your sequencer (e.g. EmailBison). [CAMPAIGN NAME] is the label for the new campaign. [PATH TO CSV OR LEAD LIST REFERENCE] is your lead list location. the sequence fields take your subject line plus email 1 and email 2 bodies. [USE DEFAULT SENDER ACCOUNTS] and [USE DEFAULT SCHEDULE] reference your standing settings.

What "good" looks like. we loaded a "YouTube test" campaign in a live test using 2 sample leads. we gave it a subject line, email 1, email 2, and the schedule plus sender accounts pre-set as defaults. a few minutes later the campaign was created in the sequencer, leads attached, sequence loaded, settings preconfigured, ready to launch.

this adds up fast. once the load workflow works, the next iteration is pointing Claude Code at lead lists you haven't touched in 3 months, having it re-verify the emails with Million Verifier or Bouncer, and auto-loading the cleaned list into a fresh campaign with your top performing angle baked in. we mentions this as the obvious next step in our test.

Use Case 3: Generating Offers

The problem. your copy is only as strong as the offer behind it. you can write the cleanest email in the world, but if the offer doesn't speak directly to what your prospect is feeling right now, they aren't replying. most agency founders fall into one of two traps. they reuse the same offer for every campaign without testing it. or they spend way too much time trying to reverse-engineer what their ICP actually cares about. either way, the emails don't convert.

The Claude Code prompt:

Generate 3 structured offers for [CLIENT NAME] targeting [TARGET INDUSTRY OR PERSONA].

Client website: [CLIENT WEBSITE URL]
Persona context: [BRIEF PERSONA NOTES: pain, goals, what's keeping them up at night]

Step 1: Run market research:
- What are the [TARGET PERSONA]'s biggest pain points right now?
- What is the dream outcome they want?
- What has failed them in the past?
- One key market insight specific to this persona
- Reference any past campaigns we've run that share patterns with this persona

Step 2: Output 3 offers. For each offer return:
- Offer name
- Front-end offer copy (one paragraph, plain language)
- The value mechanism (what the offer does for them)
- A walk-away asset they keep whether or not they buy
- Why this offer fits this persona right now

Ground every offer in the persona's actual pain and the patterns in past campaigns. No generic positioning.

How to run it. pre-load Claude Code with the context it needs. that means frameworks you trust (Alex Hormozi, Eugene Schwartz, and other names whose offer work you've vetted), your sequencer API for past winning campaigns, and any market research you've already collected on the persona. then run the prompt. you'll get a market-research pass first, followed by 3 distinct offer angles built from that pass. more context loaded into Claude Code means sharper offers on the way out.

Edit slots. [CLIENT NAME] is the company you're pitching for. [TARGET INDUSTRY OR PERSONA] is the segment you're targeting (our demo run targeted enterprise sales leaders). [CLIENT WEBSITE URL] gives Claude Code the site to crawl for context. [BRIEF PERSONA NOTES] captures everything you already know about the persona's world.

What "good" looks like. we ran this for a HubSpot-style outbound play against enterprise sales leaders. the output came back with three distinct offers.

the first was a tool consolidation play: "I'll pull your current sales tech stack, every tool that your reps touch daily and show you exactly where you're paying twice for the same functionality." the walk-away asset was a one-page consolidation map the prospect keeps regardless of whether they sign.

the second was a ramp-time killer: "let me show you what your top three reps actually do differently, discovery flow, deal progression... show you exactly where your average reps diverge."

the third was a pipeline-scoring offer that surfaced risk factors in the prospect's open deals.

each offer carried a discrete value mechanism plus a walk-away asset the prospect keeps regardless of whether they sign on. steal the structure for your next campaign.

Use Case 4: Writing Email Copy

The problem. writing copy manually means starting from scratch almost every new campaign. even when you're good at it, you spend hours writing and testing and iterating on a single sequence. the old AI options weren't much better. you'd either use Clay (a ton of steps and a ton of credits) or a basic ChatGPT prompt that had zero context on what's worked for you before. either way you ended up with generic slop that didn't match the angles your account had already proven.

The Claude Code prompt:

Write 9 cold email scripts for [CLIENT NAME] targeting [TARGET PERSONA].

Inputs:
- ICP data: [ICP NOTES OR REFERENCE FILE]
- Offers (from the offers prompt): [LIST 3 OFFERS]
- Past winning campaigns: pull top-performing copy for adjacent verticals from the [SEQUENCER NAME] API
- Tone: [TONE GUIDELINES: e.g. casual, direct, low-jargon]

Process:
1. Identify key patterns from past winners (top framework, top offer types, average word count, top subject line patterns, anti-patterns to avoid).
2. Build 9 scripts total: 3 angles per offer.
3. AB test across angles by alternating offer-first vs pain-first openers.
4. Spin tax the email body for variation (multiple phrasings per line so we don't hit spam traps).
5. Run every script through a spam filter check and flag any words that get auto-routed to promotions or filtered.
6. Return scripts grouped by offer, with subject line + email 1 body for each.

Do NOT use jargon (synergy, optimize, circle back). Keep emails short. One CTA per email. Soft CTAs work: "would that be worth a look?" beats "book a call."

How to run it. the copy workflow chains directly off use case 3. feed the offers and market research output into this prompt. Claude Code has access to your sequencer API so it pulls past winning copy patterns from adjacent verticals automatically. layer a spam-guard prompt on top so the final copy doesn't trigger filter words. expected output: 9 scripts, grouped 3 per offer, each one carrying the patterns from your historical winners.

Edit slots. [CLIENT NAME] and [TARGET PERSONA] mirror the inputs from use case 3. [ICP NOTES OR REFERENCE FILE] is your persona doc. [LIST 3 OFFERS] is the output from use case 3. [SEQUENCER NAME] is the API Claude Code pulls winning copy from. [TONE GUIDELINES] defines the voice rules for the brand.

What "good" looks like. our run on the HubSpot enterprise sales leader brief produced a subject line of "quick audit" or "your stack review" with this opener:

[FIRST NAME], if I could pinpoint exactly where your revenue team's tech stack is duplicating functionality and quietly draining rep selling time, and hand you over a one-page consolidation map in 20 minutes, would that be worth a look?

the script reads short and specific, lands a soft CTA, and handles the "am I going to get sold?" objection by promising the walk-away asset up front. the other 8 scripts followed the same shape across the other two offers. one example: a pipeline-scoring script that opens with "if I could score every opportunity in your pipeline against six risk factors (stakeholder depth, timeline compression, champion strength) and show you exactly which ones are real versus dead weight in 25 minutes, would that be worth the conversation?"

the consistent pattern across all 9 scripts: short emails, specific value, one soft CTA per email, zero "book a 15-minute call" energy. run the structure on your next batch and see what lands.

Use Case 5: Lead List Building

The problem. Apollo got most agencies started, but the data and filter accuracy were always limiting. Clay was an upgrade, and the enrichment power is real, but Clay costs serious money and still requires a meaningful setup just to get usable output. for most agency founders, a full afternoon of back-and-forth between tools is the cost of one targeted list. Claude Code closes that loop, because it connects to your stack and runs the full sequence (pull from databases, scrape communities, verify emails, enrich leads, push to the sequencer) in one workflow without paying credits per row.

The Claude Code prompt:

Build a lead list using the [AI ARC CLI] (or equivalent CLI connected to Claude Code).

Target profile:
- Job titles: [LIST OF TITLES: e.g. VP of Sales, Director of Sales]
- Company size: [SIZE FILTER: e.g. 500+ employees]
- Industry / vertical: [INDUSTRY]
- Tech stack signal: [TOOL THEY USE: e.g. Salesforce]
- Location: [GEO FILTER]
- Email verification: trust emails only (run verified email discovery)

Steps:
1. Pull the top 50 results matching the filters.
2. For each prospect return: first name, last name, headline, title, LinkedIn URL, city, full position history, LinkedIn badges, company info, company financials.
3. Flag any results that look off-target (titles or company sizes that don't match the filters) and explain why.
4. Ask me clarifying questions if any filter is too broad (e.g. "sales leader" is too vague, should we narrow to VP of Sales + Director of Sales?).

How to run it. load your prospecting CLI (we use AI Arc) into Claude Code, drop the prompt, and swap brackets for your ICP filters. Claude Code will actively push back on vague filters before it runs anything. it'll ask you to narrow "sales leader" to specific titles, or to confirm your tech stack signals. once the filters are tight, it returns the top 50 prospects plus flagged observations on which leads look off-target. from there you refine the filters, export the cleaned list, and push it to Clay if you need extra enrichment or straight to the sequencer.

Edit slots. [AI ARC CLI] is your prospecting tool's CLI. [LIST OF TITLES] should hold exact titles, not "sales leader" (Claude Code will push back on that). [SIZE FILTER] is your employee count threshold. [INDUSTRY] is the vertical. [TOOL THEY USE] is the tech-stack signal for relevance. [GEO FILTER] is the location.

What "good" looks like. our live run opened with "we're looking for sales leaders at enterprise-size companies." Claude Code immediately pushed back. "sales leader" was too broad. it asked us to narrow to VP of Sales plus Director of Sales, 500+ employees, trust emails, using Salesforce. once the filters were tight, the output came back with the top 50 prospects: first name, last name, title, LinkedIn URL, city, position history, badges, company info, and company financials. Claude Code also flagged a handful of results that looked off-target so we could refine before exporting.

the iteration loop is the whole point. you keep refining filters until the sample looks clean, then you export to your enrichment tool or directly to the sequencer. bookmark this for your next list build.

Use Case 6: Post-Booking Automation

The problem. most agency founders don't think about what happens after a call books. the reply comes in, the calendar invite goes out, the lead is confirmed, and then no automated workflow runs. you have to go into your CRM and update the deal stage by hand. someone has to send a pre-call task reminder. someone has to create the contract. someone has to log the contacts from the email thread, every single time. at 50 to 100 calls a month, that admin work piles up fast, and your team ends up doing data entry instead of closing deals.

The Claude Code prompt:

Build an n8n workflow that fires off a [CALENDLY] webhook and runs the following post-booking sequence.

Integrations needed:
- [CALENDLY] (or your calendar tool)
- [CRM NAME] (we use Attio)
- [AGREEMENT TOOL] (we use Pipedrive for agreements)
- [FIREFLIES] for call transcripts
- [SLACK] for team notifications

Workflow steps:
1. When the [CALENDLY] webhook fires, extract: contact name, email, deal/company name, meeting time.
2. Insert the contact into [CRM NAME].
3. Create a deal in the pipeline. Set stage to "discovery call booked."
4. Generate a task list for the deal:
   - Send pre-call text the morning of the call
   - Connect with the prospect on LinkedIn
   - Post-call follow-up: send call notes and recording
   - Book a follow-up call
5. When a follow-up call is booked, move the deal from "discovery call booked" to "strategy call booked."
6. Push a Slack notification to [#CHANNEL] every time a stage updates, with the prospect's info.

Separate workflow for agreements:
- When a deal is dragged from "meeting booked" or "strategy call booked" to "agreement sent":
  - Pull the Fireflies transcript and notes from the call
  - Update the agreement template based on the terms agreed on the call
  - Draft the agreement in [AGREEMENT TOOL]
  - Post a Slack notification with the draft for review before sending

Return the n8n workflow JSON so I can import it.

How to run it. this use case flips the model. Claude Code becomes the workflow builder rather than just a query runner. you describe the workflow in plain language, provide API keys for each tool, and Claude Code constructs the n8n workflow JSON for you. we didn't know n8n when we started this build. we prompted Claude Code based on the workflow we wanted, the JSON came back, and we imported it straight into n8n. expected output is a working n8n workflow that fires on calendar bookings, updates your CRM, creates the task list, drafts agreements off call transcripts, and pings your team in Slack.

Edit slots. [CALENDLY] is your calendar tool. [CRM NAME] is your CRM (we use Attio). [AGREEMENT TOOL] is your contract platform (we use Pipedrive). [FIREFLIES] is your call transcript tool. [SLACK] and [#CHANNEL] define your team notification channel.

What "good" looks like. our live setup runs like this. the Calendly webhook fires, and the contact plus deal data gets extracted automatically. the deal gets inserted into Attio with the stage set to "discovery call booked." tasks are auto-created on the deal: pre-call text, LinkedIn connect, post-call notes, and a follow-up call task. when the follow-up call books, the deal stage moves to "strategy call booked" without anyone touching it. when the deal moves to "agreement sent," the second workflow fires. it pulls the Fireflies transcript, updates the agreement template based on the terms agreed on the call, drafts the contract in Pipedrive, and posts the draft to Slack for review.

this workflow ties the first five use cases together. once the post-booking layer is wired up, the only thing left for a human is the conversation itself.

How to Combine These

the 6 workflows aren't independent. they stack on top of each other. here are 3 chains worth running once you've executed each one solo.

Combo 1. Weekly self-refining outbound loop. run use case 1 (analytics) every Monday to find the top performing campaign from last week. take that winning angle and feed it into use case 3 (offers) to generate 3 new offer variants off the same proven approach. feed those offers into use case 4 (copy) and have it write 9 fresh scripts. push the scripts into use case 2 (load campaign) so they load with your default sender accounts and schedule. launch the campaign on Tuesday. each week the campaigns get sharper because the system is training on what's already proven to work in your account.

Combo 2. Refresh dormant lead lists. run use case 5 (lead list building) against lists you haven't reached out to in 3 months and build a freshly filtered cut. re-verify the new list with Million Verifier or Bouncer to clean out the dead emails. run use case 4 (copy) to write fresh angles for the segment. then run use case 2 (load campaign) to auto-load the campaign with top performing copy pulled from use case 1's report. you get a clean re-engagement campaign in an afternoon instead of a full week of manual list-building.

Combo 3. Full post-booking flow without an ops hire. when use case 6 (post-booking automation) fires, it pulls Fireflies transcripts. extend the workflow by feeding the transcript back into use case 3 to refine the offer based on what the prospect actually said on the call. then push that refined offer into use case 4 to write a follow-up email matching the prospect's call language. your follow-ups get sharper after every call because the system is learning from each conversation you have.

Why this works. context carries through every step. analytics feeds offers, offers feed copy, copy feeds list-building, list-building feeds campaign loading, and campaign loading feeds post-booking. each step inherits the context from the step before it. run them in isolation and they save you hours. chain them together and you replace headcount.