Personalization at Scale (Clay & Claygents) prompts
Claygent and Clay AI prompts that research, score, classify, and personalize thousands of accounts — with strict output contracts built for pipelines.
51 prompts · paste-ready, no variables to fill · free
These prompts are built for the messiest part of outbound: turning raw account data into personalization that actually reads like research. Most of them run unattended inside a Clay AI column or Claygent — scraping websites, mining careers pages and case studies, detecting tech stacks, normalizing job titles, classifying accounts against your ICP, and writing first lines from whatever the scrape returned. Each pipeline prompt ends with a strict output contract: exact format, one answer, and NOT_FOUND instead of a guess, so a hallucinated detail never reaches a prospect's inbox.
They're written for RevOps builders, SDR teams, and founders running Clay, Smartlead, or Instantly at volume. Start with the scrape-to-one-liner and first-line prompts, add the QA and relevance-scoring prompts as gates, and use the strategy prompts to decide which signals deserve credits in the first place.
Build a 10-minute company deep-dive brief before you touch an account
Produces a one-page, source-backed account brief with a fixed six-section structure: snapshot, revenue logic, recent momentum, evidence-based pressures, two outreach angles, and landmines to avoid. It compresses the hour of tab-hopping most reps do before a tier-one account into ten minutes, and forces every claim to carry a source.
Generate a cold email first line from scraped research data
The workhorse of AI-personalized outbound: converts whatever research your enrichment waterfall produced into one 12–22-word observation-led first line per row. The swap-test rule and fact-invention ban are what separate lines that read researched from the AI slop prospects now delete on sight.
Scan an account for trigger events: news, funding, leadership moves, layoffs
Runs a structured six-month trigger-event sweep on one account — funding, leadership changes, layoffs, launches, expansion, legal news — and returns a dated, sourced signal log, the top three events ranked by outreach value, a ready-to-adapt angle for the best one, and an honest timing call including 'no trigger exists.' It turns 'why now' from a guess into evidence.
Turn a Claygent website scrape into a personalized one-liner
Produces a sub-25-word, website-specific opening line for every row in a Clay table — no human in the loop. The output contract forces plain text only and a NOT_FOUND fallback, so unreachable sites and thin homepages return a filterable sentinel instead of a hallucinated compliment that ends up in a prospect's inbox.
Write a connection request note (and decide if you should send one at all)
Settles the with-note or without-note question for your specific situation instead of applying a blanket rule, then writes two sub-280-character note variants with a real reason to connect and zero pitch. You get a defensible strategy call plus copy to test, not just another template.
Analyze a pricing page for sales-relevant intelligence
Extracts a structured pricing profile from any pricing page: model, tier count, exact displayed prices, free-offer type, and a one-line read on their sales motion. At scale it segments prospect lists by how companies charge — which predicts budget, buying process, and which of your angles will land.
Audit a batch of AI first lines before the campaign sends
Catches the failure modes that only show up across a batch — recycled constructions, swap-test failures, unverifiable claims — before prospects see them. You get a grade, pattern counts, individually flagged lines, and three paste-ready fixes for the generating prompt, turning QA from a vibe check into a repeatable gate.
Choose which personalization signals are worth your credits
Turns signal selection from guesswork into portfolio math: which 3–5 signals fit your ICP, offer, and budget, in what waterfall order, with what fallbacks — plus an explicit skip list and a 200-row test plan. It's the planning session that should precede every prompt-heavy Clay build, done in twenty minutes.
Classify an account into one of your defined segments
Assigns every row in your table to exactly one of your named segments, with confidence and a short evidence trail. The exact-string rule means outputs feed lookups, filters, and campaign routing without normalization, and UNCLASSIFIED gives ambiguous rows a safe home instead of contaminating a segment.
Clean up company names for email merge fields
Converts raw vendor and CRM company names into the natural short form that reads like a human typed it — no 'LLC', no ALL CAPS, no taglines — so 'quick question about {company}' merges cleanly. One of the highest ROI-per-token prompts in outbound: every recipient sees this field.
Decide which enrichment result wins in a waterfall
Resolves provider disagreements in your enrichment waterfall with explicit arbitration logic: verification over popularity, recency over rank, and a hard ban on blended values. Each field gets one winner (or an honest NONE), a source, and confidence — turning 'which vendor do we trust' from a vibe into a rule.
Design QA guardrails for AI-generated outreach before it sends
Builds a pre-send QA layer for AI outreach: automated factual, personalization, format, and compliance checks with detection logic written out, a volume-appropriate human sampling plan, a severity ladder, an automatic kill switch, and a weekly feedback loop that turns caught failures into permanent prompt fixes.
Design a prompt chain for a multi-step research pipeline
Designs a production research pipeline as a chain of single-job prompts with strict JSON contracts, validation gates, model-tier assignments, and cost math — then writes every step prompt for unattended execution. It converts the flaky mega-prompt most teams start with into a debuggable assembly line where failures surface at gates instead of polluting output.
Detect a company's tech stack from its website
Builds an evidence-backed technographic profile per account: up to eight tools with category, the exact place on the site each was spotted, and a uses-versus-integrates flag. Cheaper than technographic data vendors for targeted lists, and the evidence field makes every entry auditable and quotable in copy.
Extract a brand voice guide from your five best-performing emails
Reverse-engineers a usable voice guide from the five emails that already earned replies — fingerprint, vocabulary map, evidence-cited tone rules, a 'never' list, and a paste-ready system prompt block for AI drafting. Every rule cites your actual copy, so the guide describes the voice that works instead of the voice a workshop wished you had.
Extract hiring signals from a careers page with Claygent
Converts any careers page into a structured hiring-signal record: open role count, departments, up to three notable titles, a GTM-hiring boolean, and a one-sentence signal summary. Clean JSON per row means you can filter, score, and merge hiring triggers into copy without parsing free text.
Find a company's most recent news or press mention
Finds one verified, recent, sales-usable news item per account and returns it as structured JSON with date, source, URL, and a merge-ready one-liner. The identity check and 120-day window prevent the two classic failures: referencing another company's news, or congratulating someone on a two-year-old raise.
Find a podcast or YouTube appearance for an executive
Surfaces one verified podcast, YouTube, or conference appearance per executive with URL, date, and the stated topic — the raw material for the strongest opener in cold outreach. The triple identity check (name, company, title) makes it safe to run across thousands of prospects unattended.
Find the outreach angle hidden in a company's recent news
Works recent company news through a four-step chain — event, consequence, internal pressure, bridge to your offer — instead of stopping at 'congrats on the news.' You get dated and sourced usable items, fully reasoned angle chains, three distinct openers for the best one, and a skip list of items that would hurt to reference.
Generate 10 cold email opening lines from one piece of research
Turns a single piece of prospect research into 10 labeled opening lines, each under 25 words and each connecting an observation to a business problem. You get angle variety on demand instead of staring at a blank compose window, plus the model's pick of the two strongest lines with reasoning.
Generate LinkedIn ad copy variants for a cold B2B audience
Generates six LinkedIn ad variants across distinct psychological angles — pain, proof number, contrarian, status-quo cost, social proof, and native-style — each with truncation-safe intro text, headline, visual direction, and CTA button. You get a ranked launch order instead of six flavors of the same value prop.
Generate realistic story data for your demo environment
Generates a coherent demo dataset — realistic records in your prospect's industry language, deliberately planted records that make each demo beat land, background texture so the plants don't stand alone, and living timestamps. Replaces 'Test Company 123' with set design that lets prospects see themselves in the product.
Match your cold email to a prospect's writing style
Profiles a prospect's public writing across five stylistic dimensions, then writes your cold email in their register — their rhythm and formality carrying your offer and proof. The calibration note shows exactly where it mirrored and where it held back, keeping the technique on the rapport side of the creepy line.
Mine G2 and review sites for competitor pain quotes
Extracts up to five verbatim pain quotes from a product's critical reviews, each categorized and tagged with the reviewer's segment, plus the dominant pain theme. Feed it your competitors' G2 pages and you get displacement copy written by their own unhappy customers — the exact words your prospects would use.
Mine a local business's website for outreach signals
Builds a structured profile of any local business from its website: type, service area, owner name, tenure claim, self-stated differentiator, booking capability, and one outreach hook. It gives local-market campaigns the specificity that owner-operators respond to, at a volume no rep could research by hand.
Mine case-study pages for named customers and results
Turns any customers or case-study page into structured proof data: up to three named customers with their claimed results, plus a visible-customer count. Use it to personalize with 'you already serve teams like X', to map a competitor's install base, or to score accounts by proof maturity.
Normalize messy job titles into seniority and function
Turns the chaos of raw titles — emojis, slashes, vanity titles, abbreviations — into two controlled-vocabulary fields plus a decision-maker flag. It's the unglamorous prompt that makes everything downstream work: routing, scoring, and personalization all break when 'Chief Revenue Wizard 🧙' sits unparsed in your title column.
Personalize one demo script for every stakeholder in the room
Maps every demo attendee to their silent question, their high-relevance moments with name-addressed talk lines, their skepticism trigger with a preemption, and the one sentence you want each repeating internally. Turns a single demo flow into a call where every stakeholder feels the demo was built for them.
Personalize your call script from raw account data in seconds
Converts each pasted account record into four things: an opener insert, a problem-statement insert, the right proof story to cite, and a skip flag for bad-fit accounts, each under 25 words. It gives you Clay-grade personalization at call speed without rewriting your script per prospect, which is what keeps daily dial volume intact.
Pull verbatim prospect quotes to power your follow-up email
Mines a transcript for the prospect's exact words — pains, goals, deadlines, politics, objections — and sorts them by whether it's safe to quote them back, paraphrase them, or just privately know them. You get a ready quote bank plus the three strongest lines for your follow-up, deployed naturally.
Re-engage prospects who went dark 90 days ago
Builds a 3-touch sequence that revives 90-day-dark prospects by leading with what's new instead of apologizing for the gap. It uses their original interest as quiet context, adds proof mid-sequence, and closes with a question that's easy to answer either way — turning a dead list segment back into working pipeline.
Reply to a prospect's post, then pivot the conversation to a DM
Choreographs the comment-to-DM pivot end to end: a substantive comment, a timing rule keyed to how the prospect responds, a DM that continues the public thread instead of restarting cold, and the bridge line that eventually links the conversation to your work. The two classic failure modes — too fast and no bridge — are designed out.
Run a yes/no ICP match classifier on any account
Puts a strict tri-state gate (YES / NO / UNKNOWN) at the front of your pipeline: accounts fully matching your written ICP proceed, confirmed misfits exit with the failed criterion named, and thin-data rows queue for enrichment instead of being silently misjudged. The named-criteria output makes every rejection auditable.
Score a scraped snippet's relevance before it reaches copy
Adds a QA gate between scraping and writing: every snippet gets a 1–5 usability score, the single best fact extracted, and a reason. Gating your first-line prompt on score 4+ is the difference between personalization built on real research and confident emails built on cookie banners.
Script a LinkedIn video DM prospects actually watch
Packages the full video DM play: a one-take script under 60 seconds that opens on the prospect's world, a personalized-thumbnail move that earns the click, the companion text for non-watchers, and a re-record rule that stops perfectionism. Video attaches your face to your outreach — this makes sure it's attached to something specific.
Script a LinkedIn voice note DM that doesn't get skipped
Scripts a sub-40-second voice note engineered for the ear, not the page: name and specific reason up front, one line of relevance, and a question answerable by text. Includes the companion text message for prospects who never press play and a 15-second fallback — a rare-channel advantage most sellers waste by rambling.
Summarize a company's recent LinkedIn posts into themes
Compresses a scraped batch of LinkedIn company posts into three supported themes, a summary of the newest post, the best-performing topic, and one merge-ready outreach hook. It turns the noisiest personalization source — social content — into a structured field you can filter on and drop into copy.
Synthesize multi-source research into a 3-bullet account brief
Collapses your whole enrichment waterfall into a fifteen-second read: what the account is, what's happening there now, and the best outreach angle with its evidence. It's the bridge between automated research and human selling — reps get synthesis, not seven columns of raw scrape output to decode.
Tier your accounts into A/B/C with plays for each tier
Builds a complete account tiering system: definitions, effort allocation, channel mix, and touch counts per tier, plus promotion/demotion triggers and the capacity math to prove your team can actually run it. The output is a one-page operating system for outbound effort, not a labeling exercise.
Turn one case study into 5 segment-specific angles
Multiplies your best proof asset by five: each angle re-frames the same defensible facts for a different segment's actual concern, with a ready cold email line and the objection it quietly answers. It stops the one-size-fits-nobody case study line and stays strictly inside what you can prove.
Turn your LinkedIn profile viewers into conversations
Converts the 'who viewed your profile' list from vanity metric to warm pipeline: a triage filter for which viewers deserve action, DM and connection-note templates that never mention the view, traceback logic for guessing what brought them, and a 10-minute twice-weekly routine. The view tells you who; it must never be your stated why.
Write a bump email from a prospect's recent activity
Turns a prospect's recent post, comment, or company signal into a bump email that reacts with substance — a twist, a data point, a counterpoint — then bridges lightly back to your thread. It also scores whether the signal is strong enough to use, protecting you from the weak-signal bumps that read as stalking.
Write a cold email triggered by a job posting
Converts a job posting into a diagnosis of the initiative behind it, then into a sub-90-word email that speaks to that initiative rather than reciting 'saw you're hiring'. It forces you to pick the honest angle — make the hire succeed, ramp faster, or become unnecessary — and refuses forced connections.
Write a personalized opener from a trigger event
Converts trigger events into two-sentence openers built on the event-to-consequence structure that separates trigger-based outbound from the congratulations flood. Role-aware consequence logic means the same funding round produces different openers for a CFO and a VP of Sales — automatically, per row.
Write congratulations messages that don't feel hollow (job change, promotion, funding)
Writes trigger-event messages that survive the flood of auto-generated congratulations: specific to what the change means, carrying a no-strings offer or a sharp question, and deliberately pitch-free. Includes the follow-up window per trigger type — because the congratulations builds the door, and the follow-up weeks later is when you walk through it.
Write outreach to attendees of an event, webinar, or conference
Produces a full event outreach kit — pre-event connection notes, post-event DMs built on actual session content, a follow-up, and a message for registered no-shows. It refuses to write attendance-scraper spam: if you can't name one specific thing from the event, it sends you back to the recording first.
Write outreach to members of a LinkedIn group you share
Turns shared group membership from a lazy opener into a real channel: a filter for which members are worth messaging, separate messages for active versus quiet members, and an escalation path. If you've never participated in the group yourself, it makes you post first — visible members convert far better than lurkers.
Write personalized first lines for an entire prospect list
Processes a pasted prospect list and returns a spreadsheet-ready table of first lines, each under 22 words and built from that row's data. Weak rows get flagged instead of padded with filler, so you know exactly which prospects need more research before they enter your sequence.
Write stakeholder talk tracks: CFO vs end user vs IT
Produces three genuinely distinct talk-track cards — CFO, end user, and IT by default — each with a 60-second conversational pitch, vocabulary to use and avoid, the role's signature objection, and the type of proof that actually moves them. Ends the one-pitch-fits-all habit that quietly loses two of every three stakeholders.
Write the first DM after a connection is accepted (without pitch-slapping)
Produces a sub-50-word opener that starts a real conversation with a new connection instead of triggering the mute button, plus the reply-to-their-reply message so you're not improvising when it works. The interview forces you to articulate a reason to talk that isn't your pitch — which is the whole game.
Write trigger-based follow-ups for funding, job changes, and launches
Turns a trigger event into a follow-up that leads with what the event changes for the prospect, not a congratulation — plus two variants adapted to your other common triggers, giving you a reusable playbook. It also refuses to force a connection when the trigger doesn't genuinely map to your offer.
Frequently asked questions
A Claygent is Clay's AI web agent — it browses pages and returns answers into a table column, once per row, with nobody watching. So a Claygent prompt can't ask you questions mid-run. It has to define the task, the inputs it will receive from your columns, and a strict output contract: exact format, what to return when data is missing, and a ban on guessing. Every pipeline prompt in this category is written that way.
More prompt categories
Want this running on autopilot instead?
We wire prompts like these into a full outbound system — data, personalization, sending, and replies orchestrated end to end. You own the infrastructure. We guarantee the results.
Apply for a Pilot Spot →