Learn to build systems,
not just campaigns.

REGEX_REPLACE(TRIM({{Company}}), "(?i)[,.]?\\s*(inc|llc|l\\.l\\.c|ltd|limited|gmbh|b\\.v|s\\.a|pty|co)\\.?$", "")

IF(REGEX_MATCH({{Email}}, "(?i)@(gmail|yahoo|outlook|hotmail|icloud|proton|aol)\\."), "", LOWER(SPLIT({{Email}}, "@")[1]))

LOWER(REGEX_REPLACE(REGEX_REPLACE(TRIM({{Website}}), "(?i)^https?://", ""), "(?i)^www\\.|/.*$", ""))

IF({{Headcount}} = "", "Unknown", IF({{Headcount}} < 50, "SMB", IF({{Headcount}} < 500, "Mid-Market", "Enterprise")))

CONCAT(UPPER(SUBSTRING(TRIM({{First Name}}), 1, 1)), LOWER(SUBSTRING(TRIM({{First Name}}), 2)))

IF({{Signal Date}} = "", 9999, DATEDIF({{Signal Date}}, NOW(), "days"))

REGEX_EXTRACT(LOWER(TRIM({{LinkedIn URL}})), "linkedin\\.com/(?:in|company)/([^/?#]+)")

LOWER(CONCAT(TRIM({{First Name}}), ".", TRIM({{Last Name}}), "@", {{Company domain}}))

ROUND(
  (IF({{Headcount}} >= 50 AND {{Headcount}} <= 1000, 30, IF({{Headcount}} > 1000, 18, 10)))
+ (IF({{Last Funding Date}} = "", 0, IF(DATEDIF({{Last Funding Date}}, NOW(), "days") < 365, 25, 10)))
+ (MIN({{Tech Match Count}}, 5) * 5)
+ (IF({{Open Roles}} >= 3, 20, IF({{Open Roles}} >= 1, 10, 0)))
, 0)

IF(
  AND(
    {{Cheap Provider Email}} = "",
    {{Company domain}} != "",
    {{Headcount}} >= 25
  ),
  "RUN",
  "SKIP"
)

LOWER(CONCAT(
  IF({{Company domain}} != "", {{Company domain}},
     REGEX_REPLACE({{Company (clean)}}, "[^A-Za-z0-9]", "")),
  "|",
  SUBSTRING(TRIM({{Country}}), 1, 2)
))

You are a B2B title normalizer.

Map the title below to exactly ONE of: Founder, Sales Leader, Marketing Leader, RevOps, Engineer, Finance, Other.

Title: {{Job Title}}

Rules:
- Never invent a label outside the list.
- "Head of Growth" -> Marketing Leader. "SDR Manager" -> Sales Leader.
- If the title is empty or unreadable, use Other with confidence 0.

Return JSON only: {"persona": "...", "confidence": 0.0-1.0}

Using ONLY the homepage copy below, write one opener of max 18 words that references a specific product, customer or claim.

Homepage copy:
{{Homepage Text}}

Rules:
- No adjectives like exciting, innovative, cutting-edge, leading.
- No compliments. No questions. Plain sentence, no emoji.
- If the copy is generic boilerplate with nothing specific, return exactly: SKIP

Return the sentence only.

Read the website text and summarize the company in one sentence of the form:
"<what they sell> for <who buys it>".

Website text:
{{Website Text}}

Rules:
- Max 20 words. No marketing language. No company name repetition.
- If you cannot tell what they sell, return: UNCLEAR

Return the sentence only.

Extract every named software product mentioned in the text below.

Text:
{{Job Description Text}}

Rules:
- Named products only (Salesforce, HubSpot, Snowflake). Exclude generic terms (CRM, database, cloud).
- Deduplicate. Preserve official capitalization.

Return JSON only: {"tools": ["..."]}

Classify this email reply into exactly one intent: interested, not_now, referral, not_a_fit, unsubscribe, auto_reply.

Reply:
{{Reply Body}}

Rules:
- Out-of-office and bounce notices are auto_reply.
- "Talk to my colleague" is referral even if the tone is positive.
- Any removal request is unsubscribe, regardless of tone.

Return JSON only: {"intent": "...", "referral_name": "" }

You are a GTM researcher. From the open roles below, infer the single most likely operational pain this team is hiring to solve.

Open roles:
{{Open Roles Text}}

Steps:
1. Identify the function hiring most aggressively.
2. Find phrases describing process gaps, scale problems or manual work.
3. State the pain in one sentence a buyer would agree with out loud.

Hard rules:
- Every claim must be supported by a verbatim quote from the text.
- No speculation beyond what the postings say. If evidence is thin, set confidence low.

Return JSON only:
{"pain": "...", "function": "...", "evidence": ["verbatim quote", "..."], "confidence": 0.0-1.0}

You are a data steward resolving conflicting values for the field: {{Field Name}}.

Candidates:
- Source A: {{Value A}}
- Source B: {{Value B}}
- Source C: {{Value C}}

Decision rules, in order:
1. Discard placeholders, nulls, "N/A", and obviously malformed values.
2. Prefer the most specific value (full legal name over abbreviation, exact count over range).
3. If two sources agree and one disagrees, take the majority.
4. If all disagree and none is more specific, return the value from Source A and set confidence <= 0.4.

Return JSON only:
{"value": "...", "chosen_source": "A|B|C", "reason": "one short sentence", "confidence": 0.0-1.0}

Match our proof points to this prospect's stated priorities.

Prospect context:
{{Prospect Context}}

Our proof points:
{{Our Proof Points}}

Rules:
- Return at most 2 proof points, ranked. Each must connect to something explicitly stated in the context.
- If nothing connects, return an empty array — do not force a match.
- For each, give a 12-word "so what" written in the prospect's own vocabulary.

Return JSON only:
{"matches": [{"proof_point": "...", "so_what": "...", "prospect_evidence": "verbatim quote"}]}

Write a pre-call brief from the context below. Exactly five lines, each max 15 words, no headers beyond the labels.

Context:
{{All Enriched Context}}

Format:
What they do:
Why now:
Likely owner of the problem:
Best opening angle:
Risk / disqualifier:

Rules:
- Only use facts present in the context. Write "unknown" rather than guessing.
- "Risk / disqualifier" must name a real reason this could be a bad fit.

1. Normalize inputs — proper-case name, root company domain.
2. Provider 1 (cheapest) on all rows.
3. Provider 2 only where step 2 is empty.
4. Provider 3 (premium) only where domain is set and headcount >= 25.
5. Pattern guess as the final fallback.
6. Validation pass on every result.
7. Write a status column: valid / risky / none.
8. Only rows marked valid flow to your sequencer.

1. Webhook in from your form.
2. Reject free-email domains into a self-serve nurture track.
3. Company + person enrichment on the domain.
4. ICP fit score.
5. Branch: score >= 60 -> demo queue, 30-59 -> nurture, < 30 -> disqualify.
6. Round-robin owner assignment by territory.
7. Push to CRM with the score and a reason code.

1. Filter to rows last verified over 90 days ago.
2. Re-validate email deliverability.
3. Check the company domain still resolves.
4. Re-check current employer on the person record.
5. Move mismatches to a job-change table.
6. Stamp a verified-on date on every row you touched.

1. Weekly re-scrape of target account careers + tech pages.
2. AI column extracts named tools.
3. Flag rows containing a competitor from your list.
4. Compute days since the mention first appeared.
5. Route fresh flags to the displacement sequence.

1. Source table: closed-won contacts and past champions, with LinkedIn slug as the key.
2. Monthly re-enrichment of current company + title.
3. Diff column: current company != company on record -> CHANGED.
4. Guard: ignore changes where the new company is already a customer or is out of ICP.
5. Compute days since change; only act inside a 14-90 day window (they need time to settle).
6. Enrich the new company for fit score and existing relationships.
7. Two branches: warm re-intro to the champion, and a net-new play into their former team.
8. Write the change event to a history table so you never message the same move twice.

1. Pull from 3+ sources into one staging table, tagging each row with its source.
2. Normalize: clean company name, root domain, two-letter country.
3. Build the composite dedupe key.
4. Group by key; keep the row with the most filled fields as the survivor.
5. Merge the missing fields from the discarded rows into the survivor.
6. Suppression pass: existing customers, open opportunities, do-not-contact.
7. ICP filters last (cheapest step, so it runs on the smallest set).
8. Write the final universe to a clean table with source attribution intact.

1. Cheap firmographic pass on the entire list.
2. Fit score from firmographics only.
3. Tier split: top 5% -> Tier 1, next 20% -> Tier 2, rest -> Tier 3.
4. Tier 1: deep research — full scrape, account brief, org mapping, 5-8 contacts.
5. Tier 2: standard — 2-3 contacts, one AI personalization column.
6. Tier 3: automated only — no AI columns, template sequences.
7. Monthly re-tier so accounts can move up on new signals.

Columns, in order:
1. email (input)
2. work_email = IF(CONTAINS(email,"gmail.com"),"personal","work")
3. domain = SPLIT(email,"@",2)
4. Enrich company (domain)
5. employees, industry, country
6. icp_fit = AI column, output "strong|weak|no"
7. dedupe_key = LOWER(domain)
8. owner = lookup from territory table
9. route = IF(icp_fit="strong","AE","nurture")

Columns:
1. company_domain (input)
2. Company enrichment
3. Scrape homepage + /pricing + /careers
4. Claygent: tech stack summary
5. Claygent: hiring signals (last 90 days)
6. Claygent: recent funding or launches
7. AI column: 120-word account brief citing the columns above
8. AI column: 3 opener angles, ranked

Columns:
1. company_domain (input)
2. Find people: titles = persona list, limit 3
3. Email waterfall: provider A -> B -> C, stop on first valid
4. verified = IF(email_status="valid",TRUE,FALSE)
5. LinkedIn URL enrichment (fallback path)
6. send_ready = AND(verified, NOT(ISBLANK(first_name)))
7. cost_notes column for credit tracking

Columns:
1. account_domain (input, synced from CRM)
2. Headcount now vs 90 days ago
3. headcount_delta = (now - past) / past
4. Claygent: leadership changes in your champion's function
5. Claygent: competitor logo on their site or job posts
6. risk_score = weighted sum of the three signals
7. flag = IF(risk_score > 0.6,"call this week","monitor")

Nodes:
1. Webhook (POST /clay-row) — the URL you paste into Clay's HTTP API column
2. Set — map Clay fields to CRM field names
3. IF — drop rows with no verified email
4. HTTP Request — CRM search by domain
5. Switch — found? update : create
6. HTTP Request — upsert
7. Error Trigger + Wait 60s + retry once
8. Slack — post failures to #gtm-ops

Nodes:
1. Schedule Trigger — 07:30 local, weekdays
2. HTTP Request — pull signals from the last 24h
3. Code — dedupe by domain, keep the strongest signal per account
4. Filter — ICP domains only
5. Item Lists — group by owner
6. AI node — write a 3-bullet summary per owner
7. Gmail/Slack — one message per owner
8. NoOp — log run summary

Nodes:
1. Schedule Trigger — every 15 minutes
2. Postgres/Sheets — select 50 pending rows
3. Loop Over Items — batch size 10
4. HTTP Request — Clay table create-row endpoint
5. Wait — 5s between batches
6. Update source row status = queued
7. IF non-2xx -> status = failed + Slack alert
8. Daily Schedule -> re-queue failed rows once

Nodes:
1. Webhook — form submission
2. Set — normalize email, domain, source
3. HTTP Request — push row to Clay, wait for enrichment callback
4. Webhook (callback) — receive enriched row
5. IF — icp_fit = strong?
6. AI node — draft a reply referencing their company and stated need
7. Gmail — send from the owner's mailbox
8. Slack — notify owner with the brief and the sent copy

1. Ingest the intent feed (webhook or scheduled import).
2. Resolve the account: normalize domain, match against your universe.
3. Drop unknown or out-of-ICP domains immediately.
4. Dedupe signals per account within a 7-day window.
5. Map signal type to sequence: pricing page -> direct, competitor content -> displacement, hiring -> capacity.
6. Pick contacts by the persona that matches the signal.
7. Enforce a global cadence cap so one account cannot be hit by three plays at once.
8. Log every routed signal with its outcome for weekly review.

IF(REGEX_MATCH({{Raw Phone}}, "[a-zA-Z]" ), "Invalid", CONCAT("+", {{Country Code}}, REGEX_REPLACE({{Raw Phone}}, "[^0-9]", "")))

You are an enterprise SDR. Read the following raw 10-K risk factors section.

Text:
{{10-K Text Dump}}

Rules:
- Identify EXACTLY ONE major operational or supply chain risk mentioned.
- Do not summarize. Quote the exact sentence where they state the risk.
- Write a 1-sentence 'so what' on how our software solves this.

Return JSON only: {"risk_quote": "...", "solution_angle": "..."}

The prospect replied with an objection. Draft a response.

Prospect Email: {{Prospect Email}}
Our known competitors: {{Our Competitors}}

Rules:
- Be brutally concise. Max 3 sentences.
- NEVER be defensive. Acknowledge their point immediately.
- If they mention a competitor, do not attack the competitor. Highlight our specific differentiator instead.
- End with a low-friction question, not a push for a call.

Draft the response directly. No pleasantries like 'Hi' or 'Best'.

1. Apollo: Export raw lead list based on generic filters (Title, Industry).
2. Clay: Import CSV.
3. Clay: Run Company Domain through Waterfall (Clearbit -> Prospeo -> Waterfall).
4. Clay: Verify Apollo emails via ZeroBounce API column.
5. Clay: Filter out Bounced or Catch-All emails.
6. Clay: Run Claude Personalization Prompt on remaining valid leads.
7. Webhook: Push clean, enriched, personalized rows to HubSpot CRM.

1. G2 Integration triggers when an account surges on your category.
2. Clay checks CRM: Does this account exist? Is there an open opp?
3. If open opp: Slack alert the AE. Halt enrichment.
4. If net-new: Find 3 contacts (Champion, Decision Maker, Technical Evaluator).
5. Enrich contacts via email waterfall.
6. Trigger direct mail (Sendoso) to Decision Maker.
7. Trigger cold email sequence to Champion mentioning the category research.

Nodes:
1. Webhook — Listens for Slack slash command payload.
2. Set — Extract the domain from the Slack message.
3. HTTP Request — Push domain to Clay webhook.
4. Wait — Wait for Clay callback (Webhook).
5. AI Node — Format the returned Clay data into a clean markdown brief.
6. Slack — Post the brief back to the channel as an ephemeral message.

IF(REGEX_MATCH({{Raw Revenue String}}, "(?i)B"), "Enterprise", IF(REGEX_MATCH({{Raw Revenue String}}, "(?i)[5-9]\d\s*M|\d{3}\s*M"), "Mid-Market", "SMB"))

Analyze the following recent news article about the company.

Article Text:
{{News Article Text}}

Task:
Find exactly ONE direct quote from a C-suite executive (CEO, CRO, CTO, etc.).
The quote must be related to growth, a new challenge, or a strategic pivot.

Rules:
- DO NOT summarize. Return the exact quote in quotation marks.
- Include the name and title of the person who said it.
- If there are no direct quotes from an executive, return exactly: NO_QUOTE_FOUND

Return JSON: {"executive": "...", "title": "...", "quote": "..."}

IF(OR({{Headcount}} = "", {{Headcount}} = 0), 0, ROUND(TO_NUMBER(REGEX_REPLACE({{Estimated Revenue}}, "[^0-9]", "")) / {{Headcount}}, 0))

You are a brutal, top-performing SDR manager. Rewrite this email.

Draft:
{{Draft Email}}

Rules:
- Strip all pleasantries ("Hope you are well", "Thanks for connecting").
- Delete all adjectives (innovative, leading, powerful).
- Maximum 3 sentences. Maximum 50 words total.
- Must be easily readable on an iPhone lock screen.
- End with a low-friction question, not a meeting request.

Return only the rewritten email body. No commentary.

You are preparing a rep for a discovery call. Read the prospect's background.

Background:
{{LinkedIn About Text}}
{{Recent Posts Text}}

Task:
Write exactly 3 sharp, provocative discovery questions. 
Do NOT ask things we could Google (e.g. "What are your goals?").
Ask questions that challenge their current operating model based on what they've posted.

Return JSON: {"questions": ["Q1", "Q2", "Q3"]}

Read these recent messages from a customer.

Messages:
{{Recent Communications Text}}

Task:
Analyze the tone and detect if this account is a churn risk.
Look for: radio silence after a problem, executive escalation, or repeated frustration with the same bug.

Rules:
- Rate the risk: LOW, MEDIUM, or HIGH.
- Give a 1-sentence reason.
- Do they mention a competitor? (true/false)

Return JSON: {"risk_level": "...", "reason": "...", "mentions_competitor": true/false}

Analyze this competitor's pricing page text.

Text:
{{Pricing Page Text}}

Task:
Identify exactly how they gate their pricing tiers (e.g., is it by User Seats? API calls? Monthly Revenue?).
What feature forces a buyer to upgrade to their "Enterprise" tier?

Return JSON: {"value_metric": "...", "enterprise_gate": "..."}

You are a world-class Conversion Rate Optimization (CRO) expert. Analyze this SaaS paywall copy.

Paywall Text:
{{Paywall Copy Text}}

Task:
1. Identify the biggest point of friction or ambiguity that would prevent a free user from upgrading.
2. Suggest 2 specific behavioral psychology tweaks (e.g., anchoring, scarcity, loss aversion) to apply to the copy.
3. Rewrite the main Call to Action (CTA) to be entirely value-driven instead of action-driven (e.g. not "Upgrade Now").

Return JSON: {"friction_point": "...", "psychology_tweaks": ["...", "..."], "new_cta": "..."}

You are a technical SEO strategist building a programmatic SEO (pSEO) campaign.

Product: {{Product Description}}
Audience: {{Target Audience}}

Task:
Design a programmatic SEO campaign structure.
1. Define the URL syntax pattern (e.g. /integrations/[software-a]-and-[software-b]).
2. Define the 3 dynamic data points that must be injected into every page to make it rank.
3. Provide 3 example URLs based on your pattern.

Return JSON: {"url_pattern": "...", "dynamic_variables": ["...", "...", "..."], "example_urls": ["...", "...", "..."]}

Nodes:
1. HubSpot Trigger — Listens for "Deal Created" where Amount > $10,000.
2. HTTP Request — Hits Clearbit/Apollo API with the associated Company Domain.
3. Set — Maps the enriched data (Headcount, Funding, Tech Stack).
4. Slack — Posts a beautifully formatted block-kit message pinging the assigned AE.

Nodes:
1. Gmail/IMAP Trigger — Watches for unread emails in the shared inbox.
2. OpenAI — Prompts: "Classify this email into: Pricing, Support, Spam, or Partnership".
3. Switch — Routes the workflow based on the OpenAI classification.
4. Gmail (Reply) — Auto-replies with specific FAQ templates or routes to Zendesk.

Nodes:
1. Webhook — Receives data from Apify/Phantombuster LinkedIn scraper.
2. Data Fetcher — Cleans the commenter's profile URL and Name.
3. Dropcontact / Hunter — Finds the commenter's B2B email address.
4. Outreach/Smartlead — Drops the verified email into an auto-sequence.

Nodes:
1. Stripe Trigger — Listens to 'invoice.payment_failed' events.
2. IF Node — Checks if the subscription MRR is > $500.
3. True Branch: Slack — Pings the CSM channel to manually intervene.
4. False Branch: Gmail — Sends a generic automated "Update your credit card" email.

Nodes:
1. Typeform Trigger — Fires when a new demo request is submitted.
2. Code Node — Formats the JSON answers into clean key-value pairs.
3. Notion — Creates a new page in the "Inbound Leads" database.
4. Notion — Appends a Checklist block inside the page for the SDR to follow.

1. LinkedIn Sales Nav: Build a highly targeted Boolean search (e.g. VP Marketing AND SaaS).
2. Scrupp / Phantombuster: Use the Chrome extension to extract the search results.
3. Scrupp: Export the extracted profiles as a CSV (or send via Webhook).
4. Clay: Import the CSV into a new table.
5. Clay: Run "Enrich Person from LinkedIn Profile" to grab their current Company Name and Domain.
6. Clay: Run the "Find Work Email" Waterfall integration (Prospeo -> Dropcontact -> Hunter -> Datagma).
7. Clay: Run the "Verify Email" integration to strip out Catch-Alls and Invalid emails.
8. CRM: Push the fully enriched, 100% verified list into HubSpot or Smartlead.

1. Apollo: Export leads that are missing email addresses.
2. Clay: Import the list (Name + Company).
3. Clay: Use "Search Google" integration with query: "site:linkedin.com/in/ {{Full Name}} {{Company}}".
4. Clay: Map the first Google search result URL to a new column (this is their LinkedIn Profile).
5. Clay: Pass the newly found LinkedIn URL into the Email Waterfall finder.
6. Clay: Verify the email and sync back to your sequence.