I think most people are wasting their LinkedIn outreach thinking of it like email instead of it's own channel. Here's the 3 strategies I'd do instead. I was speaking to James Winter about some suggestions I had for his portcos with Claude code and such and I brought up how one of their portcos wanted to sync LinkedIn outreach with email. "We think if we email them twice, we should move over to LinkedIn." In my opinion, that's a waste of the channel. LinkedIn outreach (especially using voice notes) 99/100 times has higher reply rates than cold email but lacks the ability to scale without fake accounts. If you only had 25-50 connection requests to send per day, why would you use the same strategy of email where the channel scales linearly. When you're doing LinkedIn outreach, I'd focus on these things first. 1) Use a tool like Trigify.io or some Rapid API endpoints to find your ICP engaging and connecting with customers you have permission to name drop. Find everyone engaging with your customer's employees posts and their company page. Don't say, "I saw you engaging with my customer's post" but instead send your regular message you usually would but leave a custom variable in your social proof line of your message to reference that customer. Instead of sending a message saying, "We help companies like {{company they have never heard of}}...." you can name drop that customer case study who they will of course know. 2) Install every website deanonymization software on your site with a free trial. if they do contact lookup, clean them up and make sure they are ICP. If they do company lookup, use Clay or Prospeo.io to find contacts. Target people and companies with a high likelihood of being on your site. Again, don't mention anything about being on the site. Just reach out like you normally would and put a little delay on it. Track which tools give you the highest reply rates/success and pay for them. Probably a whole post can be dedicated on why I suggest you try them all for your niche but it's free, so trust me on this one. 3) Reach out to people showing intent about your solution/problem/they are doing research/etc by targeting people in your ICP engaging with posts on the topic (or your competitor's post....) Same strategy here, just going a little bit more broad and using the last amount of capacity you'll have because of the limits of reaching out from 1 account. So instead of thinking about LinkedIn as an extension of email, remember you only get so many shots on goal and the reply rates are so much higher on LinkedIn. Use them to reach out to some of your hottest leads. | 22 comments on LinkedIn

$3.42 to find 49k Instagram profiles including their emails and some light ICP building ready to go? Let's break it down. Scraping Instagram is notoriously difficult but so many people want to target creators/contacts that really aren't on LinkedIn (thank goodness the LinkedIn engineers aren't as good at blocking scraping as Instagram). So we have to go to our favorite scraping partner on the internet. Hey there Google! Google has built you a search scraper that everyone company must let it in in order to be discovered. In this process, we used OpenWeb Ninja's google search api to do it. Let's try to break this down. Google scraping is great but you can only get a subset of searches at a time. So if you want a ton of contacts, you're going to have to run lots of different searches programmatically. For this search we were targeting CEOs, authors, founders, etc. Not necessarily professional content creators. I had Claude code build me searches for every state, every city, and then 200 keyword searches to pair with each base keyword. We would scrape Google searches where the search would be site:instagram.com "founder" AND "@" AND "Los Angeles" AND "Finance" vs. site:instagram.com "founder" AND "@" AND "Los Angeles" AND "MedSpa" (the @ is to have the highest chance that someone has an email in their bio!) If we only search for site:instagram.com "founder" no way we get the whole search. Then Claude code ran with the OpenWeb Ninja API and completed the searches for me while deduping in the process. Going to throw it in Clay to orchestrate some final mile Apify enrichment and email validation before launching the campaign! | 20 comments on LinkedIn

Most people start with Layer 3 (see the graphic below). They obsess over subject lines and personalization before they've even built an offer that works cold. That's backwards. There are 3 layers to outbound: → Layer 1: Outbound Strategy ICP definition, offer design, and positioning. If your offer doesn't deliver value density before the meeting, no execution will save you. If your ICP is too broad, you'll burn volume on people who'll never buy. This is where differentiation happens. Most people skip it because it's hard. → Layer 2: Outbound Marketing Plan Targeting rules, list-building logic, and channel mix. This is where strategy turns into structure. Segmentation by use case. Measurement frameworks. Qualification criteria. Without this, you're guessing at scale. → Layer 3: Execution & Scale Cold email sending, SDR workflows, and follow-up sequences. This is where everyone starts. That's why most outbound campaigns plateau early and can't scale past low volume. The reason this order matters: Skip a layer, and the next one fails. Skip Layer 1 → weak offer gets exposed at volume Skip Layer 2 → execution breaks when you try to scale Start at Layer 3 → polite declines, forever Outbound works when you treat it like engineering, not marketing. Ready to test this? Manufacturer or professional services company doing $2M–$3M+ with an LTV above $7K? Book a pilot. We'll get you 3-4 qualified meetings before you pay anything. Apply here: https://bit.ly/C17Pilot Repost this to help founders stop optimizing the wrong layer first. Follow Enzo Carasso 🧲 for more on outbound engineering, offer design, and controllable pipeline. | 104 comments on LinkedIn

The most dangerous and biggest BS lie in sales: “AI will fix our process.” It won’t unless the process is built for AI. Most teams don’t need more tools They need AI infrastructure with the inputs, signals, outputs, and systems that actually make AI useful. Here’s what a real AI-native sales engine looks like under the hood: AI Inputs → ICP, buyer context, triggers, tone AI Signals → stagnation, sentiment, leakage, engagement AI Outputs → personalization, timing, risk alerts, forecast checks AI Systems → outbound frameworks, coaching loops, pipeline intelligence When these pieces connect, AI stops being a writing assistant and starts being the operating system your revenue team runs on. Cleaner forecasts. Higher replies. Smarter coaching. Tighter pipeline visibility. DM me “AI” if you want to start working like a real SaaS Founder/CEO, using AI in the way it matters. If that’s you, then stop running your company like it’s 2022 and let’s implement a revenue machine together. 👉 Follow Alan Ruchtein for AI-native GTM systems that actually move revenue. | 76 comments on LinkedIn

I've always wondered what tools Ramp used for GTM. Turns out they use... Ramp? I saw an awesome post from Parth Gujare on X yesterday where he showed off some product screenshots of Ramp Revenue, the company's internal revenue orchestration tool. The product included signal aggregation, message drafting, data enrichment and more, all made specifically for the Ramp team. This particularly excites me because it harkens to the idea that GTM teams shouldn't be using out-of-the-box tools to get their work done. If a system isn't built completely custom for your company, it's probably not working as efficiently as it could be. For companies that don't have the internal resourcing to create their own GTM software, I've always felt that automation tools like Clay are their best bet. However, it's getting to the point where spinning up products like this doesn't take a 10 person engineering team, or really anybody technical. As the barrier to entry decreases for bespoke GTM tech, I think we'll see a lot more companies doing this. I've always felt that the Ramp GTM team is one of the best in the world, and now we know why after a look under the hood! Props to Parth for posting this, hopefully it inspires more people to take initiative over their company's stack 📚 | 28 comments on LinkedIn

Elon Musk bought X and implemented signal-based outreach to sell ads. And honestly, it’s not the worst play in the world: - The email looks human - It grabbed media from my post - They did some filtering to see I talk about business - They mention “testing” paid ads (lower-risk) - They emphasize the “free hands-on support” I’m sure there are ways they’re identifying business owners and automating a lot of the tedious work to set this up, but the end product still really isn’t bad. This is your signal that no matter how big your company, using the right signals for outreach likely isn’t a bad idea. | 34 comments on LinkedIn

Scale doesn’t break outbound. It reveals what was broken already. Most systems were never built to handle volume. You scale up, cracks appear, and you're back to manually fixing every piece instead of running an engine. Here's how to build outbound that survives scale: 1/ Account Sourcing Decide exactly which accounts you will and will not target. Most outbound fails because you're contacting people who will never buy. → Build ICP rules using firmographics and economics → Reuse the same account set instead of rotating lists → Precision at sourcing prevents waste at every step after 2/ Outbound Distribution Your message has to reach the market. Perfect targeting and a strong offer mean nothing if emails don't land. → Run high-volume sequences that maintain deliverability at scale → Keep in mind that infrastructure often breaks here first when you grow 3/ Offer-Led Framing Give them a concrete reason to respond. Precision without a superior offer leads to commoditization. → Anchor to a specific revenue problem, cost issue, or operational constraint → This is where most companies stall out 4/ Segmentation by Use Case Different buyers have different triggers. → Segment by use case and buying trigger, not persona or industry labels → One size fits all kills response rates at scale 5/ Content Reinforcement Support material validates your outbound claims. → Use case studies, teardowns, and proof points to reinforce credibility → Build a library of resources that moves deals forward 6/ Live Feedback Loops Most operators run outbound blind and wonder why results plateau. → Review replies weekly and tag patterns → Rebuild offers when data shows you're losing on value, not targeting 7/ Pre-Meeting Value Earn the meeting before discovery happens. → Deliver insight before a call is booked → The offer delivers something concrete before the first call 8/ Qualification Rules Reject meetings that fail firmographic fit, use case, or economic criteria. → If you book meetings with people who can't buy, you waste sales capacity → Qualification helps protect your time and close rates 9/ Back to Account Sourcing The flywheel repeats. Wins and losses feed back into targeting rules, offer design, and segmentation. Most outbound is built like a list of tactics. This is how you build it like an engine. Want to see it in action? We'll run a pilot campaign, book 3-4 qualified meetings, and prove the system works before you pay a dollar. Apply here: https://bit.ly/C17Pilot Repost if you're rebuilding your outbound system from scratch. Follow Enzo Carasso 🧲 for practical systems to scale your business with outbound. | 48 comments on LinkedIn

Half your lead list is just… missing (Not people, but missing emails you can actually send to) And that’s where most outbound motions quietly die… Because the typical loop looks like this: run one provider → blanks…run another → “valid” emails that still bounce clean the CRM → patch the list → repeat …and somehow we call that building pipeline 🫠 This is why i think Clay waterfalls + Icypeas is such a good combo Icypeas is built in a way that GTM teams actually use: Full Name + Company Domain → verified work email Then Clay does what it’s best at: waterfall across multiple providers until you get a real hit ... instead of betting your whole campaign on one provider. The setup is simple : Select Icypeas account → Clay-managed or your own API key Map inputs → Name + Domain (optional catch-all) Enable auto-update → refresh every 24h (less stale data) Add a condition → run only if email is blank (credit-efficient) Waterfall = higher coverage without extra tabs Verification = fewer bounces / healthier deliverability Auto-update = less stale data when titles change fast Conditional runs = you don’t pay twice for the same row Net result..more filled rows…Cleaner lists…Fewer bounces. ✌️ and reps start outreach with data that actually holds up. | 50 comments on LinkedIn

Outbound decisions are usually made with the same few words They show up long before emails go out While teams are planning, forecasting, and prioritizing These are the words that quietly guide where time goes What gets attention And what moves forward 👇A reminder of the GTM language behind those everyday conversations P.S. Might be worth saving as a quick memory refresh for later | 68 comments on LinkedIn

Bad deliverability will kill your outbound before your offer ever gets read. But everyone blames the copy. I've sent millions of cold emails. Here are the 7 deliverability mistakes that destroy campaigns: 1/ Using a Single Domain for Outbound → Running outbound from one domain concentrates all risk in one place. → If you seriously want to scale you need at least 100 domains. 2/ Sending Too Much From Each Inbox → High send counts create abnormal patterns that trigger filtering. → Cap daily sends at 15-20 emails per day and to never go above 30. 3/ Treating Warm-Up as a One-Time Step → Inbox reputation is shaped by ongoing behavior, not initial setup. → Maintain consistent sending for 2-3 weeks minimum and avoid pauses that reset trust signals. 4/ Scaling an Offer That Doesn't Get Replies → Low engagement trains filters to deprioritize your emails. → If nobody's replying, pause scaling and adjust your offer before touching infrastructure. 5/ Using Opens as a Performance Signal → Opens do not indicate intent. → Track replies and interactions at the inbox level and remove inboxes with sustained low engagement. 6/ Adding Infrastructure Before It's Required → Complex setups delay feedback instead of improving performance. → Add complexity only when manual limits are reached. 7/ Treating Deliverability as a Solved Problem → Inbox health degrades over time. → Rotate domains proactively and adjust volume based on engagement trends. Outbound isn't about clever subject lines. It's about engineering systems that survive scale. If you're doing $2M-$3M+ in revenue and want to see how we engineer outbound pipelines that actually deliver, apply for our pilot campaign. We'll book 3-4 qualified meetings in your calendar. You verify ICP fit. If they're good, we talk. Apply for a pilot campaign at https://bit.ly/C17Pilot Repost if you believe infrastructure beats tactics. Follow Enzo Carasso 🧲 for outbound engineering at scale. | 103 comments on LinkedIn

Here's how to automate 90% of your proposal generation with Clay and Google Slides: I made a video a few months ago where I used publicly available company info to auto-generate slide decks with Gamma. However, I had two small qualms with it- it was using zero first-party data, and the deck looked a little different every time. Luckily Akansh Gupta went ahead and fixed this for me with a v2. Here's what the table does: 1. Ingests a call transcript, and then enriches the person and company involved in the call 2. Uses the transcript to fill variables required by a Google Slides template we created 3. Sends the variables to a Google App Script, which duplicates our template proposal and fills it with the variable data 4. Voila, proposal complete! For anyone using Gong or other call recording software with a good API, I highly recommend giving a variant of this a try. Quick disclaimer- this was meant as an example table, it will likely look a bit different in practice as you'll want to use the webhook from the call recorder as the table source, not a "find people" integration! Quick video explainer below ⬇️

Deliverability is the hidden killer of GTM strategies. When your open rates dip, the immediate instinct is to rewrite the subject line or blame the lead list. But more often than not, it’s a plumbing issue, not a people issue. This is the specific gap Warmy.io - Email channel. Reliable. fills. It acts as a safety net for the things that keep founders and Ops leads up at night: The Blind Spots: → See your domain’s health in real time, no more guessing if SPF or DKIM broke. The Spam Jail Fear: → Simulated human engagement helps your emails land where they belong. The Ops Headache: → Bulk mailbox connections and API access make scaling painless. It’s simple: Fix the infrastructure first, and the revenue follows. Don’t let your best pitch die in the Spam folder. Try it free here: https://lnkd.in/gEq79hJP | 25 comments on LinkedIn

Most cold emails never reach the inbox. Here’s how to fix that 👇 Let's break down the 4 tiers every sender MUST master to hit 90%+ inbox placement: 1. Technical Foundation → Authenticate with SPF, DKIM, and DMARC → Diversify inboxes across Gmail, Outlook, SMTP → 3 inboxes per domain 2. Warm-Up & Volume Management → Warm up for at least 30 days to build sender reputation → Ramp up daily sends gradually → Fully ramp to 25 sends per inbox per day 3. List Perfection → Double verify your lists to remove invalid emails → Use a catch-all verifier to maximise validation coverage → Target high-quality, relevant audiences only 4. Content Strategy → Use plain text only (no HTML, links or attachments) → Use a ton of spintax → Write like a real human, not a marketer Nail all four tiers, and you’ll be in a good area. If you're still not seeing results and your inboxes appear healthy (check warm up % score is > 95% in Smartlead/Instantly) then you have an offer issue. Change up your copy and go again. StackOptimise ⚙️ | 17 comments on LinkedIn

8 AI Prompts That Actually Book Meetings (Cold Email Advanced Edition) (+ the exact wording I use to land 79+ meetings/month) ☑ ICP Research Prompt “Act as a sales analyst. Analyse CTOs in B2B SaaS (100–500 employees). Summarise their top 5 priorities, daily pains, and buying triggers in less than 200 words.” ☑ Personalization Hook Prompt “For this LinkedIn profile [link], draft 3 icebreakers. Keep under 20 words, avoid flattery, tie to their recent activity. Avoid jargon, buzzwords or being salesy.” ☑ Problem Hypothesis Prompt “Write 3 pain-point hypotheses for VPs of Sales in SaaS who missed quota last quarter. Keep each under 15 words. Make them relevant and give me a 1 line “Why it matters” for me to understand your POV” ☑ Objection Handler Prompt “Give me 3 tactical responses to ‘too expensive’ objection for SaaS buyers in X industry for Y role/title that are willing to buy [my solution]. Make it consultative, not defensive. Each reply must be under 25 words.” ☑ CTA Prompt "Generate 3 high-converting CTAs for a SaaS demo request email. Each CTA must be under 12 words, avoid pushy sales language, and emphasize clear, outcome-driven value that motivates action without pressure. Prioritize clarity, psychological pull, and alignment with a business decision-maker mindset." ☑ Follow-up Prompt "Craft a compelling two-sentence follow-up email for a prospect who opened but did not reply to my initial message [insert message 1]. Provide fresh value by weaving in a single new, research-backed insight or market trend that sparks curiosity and relevance, while keeping the tone consultative rather than transactional." ☑ Case Study Builder Prompt "Transform the following customer testimonial into a concise, two-line cold email case study. Highlight quantifiable ROI, specific numbers, and credible impact in a way that feels authoritative yet humble—avoiding overt self-promotion while ensuring the results command attention and trust." ☑ Multi-Thread Prompt "Identify 3 alternative stakeholders to approach if a VP of Sales does not respond. For each contact, justify their relevance by explaining their role in the decision-making ecosystem, their potential influence on the buying process, and how they could shape or accelerate a purchasing conversation." ♻️ Repost this so more people stop using GPT like an intern | 47 comments on LinkedIn

The time-saving shortcut I discovered for figuring out for better data ❣️ #TLRD: I kept seeing the same situation across teams: Replies were coming in here and there. But the pipeline felt inconsistent. Nothing was obviously broken (which made it worse) At first, we did what most teams do: We changed the copy Fine-tuned targeting Plus a few levers you’d do to fix things quickly… What actually worked was zooming out and looking at outbound as a system... (of course) First: where the data comes from. Then: whether emails and mobile numbers actually work….. not just “marked” valid.... but verified through an email waterfall across multiple sources BetterContact helped us here. Many teams underestimate the importance of data (because it feels “we can figure out that later” problem) but you actually can’t ⚔️ …. and once you have the right data, follow this blueprint to ACTIVATE it ✌️ 1️⃣ Where's the list coming from? Roles change fast. Are we keeping that updated? 2️⃣ Do the emails and mobiles actually work? 3️⃣ Does the data cover everything you’d need for the campaign/motion? 4️⃣ Does the outreach really make sense? Enriched data → clear signals → messaging that connects. | 67 comments on LinkedIn

Most founders stare at the wrong GTM metrics. That’s why their outbound “works”… but revenue doesn’t move. 7 KPIs that actually tell you if outbound is working 👇 KPI 1: Cost to Prove Value What it really is: how much you spend to show real value before a meeting. Why it matters: if CPV creeps toward LTV, scale just makes you poorer faster. KPI 2: Qualified Meeting Rate What it really is: % of meetings that should have been on the calendar. Why it matters: a “busy” calendar can still be 50% wasted sales time. KPI 3: Response → Meeting Conversion What it really is: how many replies turn into booked calls. Why it matters: interest without meetings means your offer doesn’t create urgency. KPI 4: Pipeline per 1,000 Contacts What it really is: dollars of qualified pipeline for every 1,000 people you touch. Why it matters: activity is a cost center; pipeline is what pays salaries. KPI 5: Time to First Value What it really is: how fast prospects feel the value of talking to you. Why it matters: early clarity shortens sales cycles. Late value kills deals. KPI 6: Disqualification Rate What it really is: how often your system says “no” before sales ever gets involved. Why it matters: if nothing is disqualified, your filters are fiction. KPI 7: Outbound Pipeline Coverage What it really is: % of next quarter’s target covered by outbound-generated pipeline. Why it matters: if outbound can’t cover targets, growth is hope, not control. These KPIs aren’t about doing more. They’re about knowing what to control. If you want to see how C17 Lab builds outbound that hits these numbers: We run a live outbound pilot, build the system, run a real campaign, and you keep the pipeline before you decide to scale with us. Comment “PILOT” and I’ll DM you the details. PS: If you know a founder still bragging about “emails sent” instead of “pipeline created,” repost this for them. | 97 comments on LinkedIn

Clay keeps upselling you to the most expensive AI models. Every time you run a Claygent enrichment, it recommends Argon or Claude Opus when you probably don't need them. This is the FASTEST way to burn through Clay credits. (ask me how I know 😂) Here's the fix: Connect Clay to OpenRouter. OpenRouter is a unified interface for LLMs that automatically selects the cheapest, most effective AI model for each task. Instead of guessing whether you need Claude for copywriting or GPT-4o Mini for basic research, OpenRouter picks the right model every time. And it keeps costs as low as possible (unlike Clay's built-in recommendations). What happens when you use it: 1. For simple tasks (counting months in a year), it routes to Mistral AI Nemo or Deep Infra Inc. 1. For Web scraping with Claygent, it automatically switches to Perplexity. 3. For copywriting, it'll use Anthropic’s Claude 3.7 Sonnet. You don't select ANYTHING. It just works. See the full tutorial below that I recorded showing how to set this up using Clay's HTTP API enrichment. Fair warning… You need the $349/month Clay plan to use HTTP API. But if you're already on that plan and burning credits on expensive models you don't need, this is a no-brainer. PS: As always - credit goes to one of my fav Unlock Clay students, Olly Sleep(tag), for shedding light on OpenRouter a while back Link to vid: https://lnkd.in/eUYiHgY9 | 24 comments on LinkedIn

This is the type of campaign I would suggest anyone runs if they are new to cold email and automated outbound. Internally, we call it the creative ideas campaign and if you watch the video (which I'm soft launching before we put it on YouTube) I go over the past and current success we've had with this campaign. If you don't have time to watch the video, the strategy is that we use AI to learn about our customer's offer and how they help their customers. So that when we reach out to a prospect, we can analyze their website and understand what they do in order to output 1-3 creative ideas on how we would be able to help them. This way we aren't sending AI generated slop compliments but instead, real tangible ways that we can help based on the company and contact's context. This is the best campaign for many of our customers including Clay, a PR company we're about to release a huge case study with that gets 80 leads a day with us, and so many of our other customers. Here's an example template that we could use Hey {{first_name}} - I had some ideas about how we could target {{customer types}} together. Here's what I'm thinking {{Idea}} {{Idea}} {{Idea}} We built a platform that lets you get started for $100 a month with everything you need to start campaigns like I mentioned. Could I show you how it works sometime? {{sendingAccountName}} Here's what that could look like if it were filled in. Hey {{first_name}} - I had some ideas about how we could target Finance Leaders at US Pharma companies together. Here's what I'm thinking - Target CFOs that recently took over their role and show them how you can help them reduce expenses in their first 90 days. - Use our Website Visitor Identification to uncover purchasing directors researching your site but didn't leave their email. - Send campaigns like this to VPs of Finance asking if their org is currently evaluating currently cutting software costs with some of the NIH cuts happening. We built a platform that lets you get started for $100 a month with everything you need to start campaigns like I mentioned. Could I show you how it works sometime? And for those that might say this is really long, I get it. You can simplify it into one sentence that applies to the prospect which we also helped 3x the reply rates for John Thomas campaigns by just adding this one line to his best templates. Check out the video or save this for later to understand more about the strategy. | 16 comments on LinkedIn

Most companies get to $2-3M ARR then stall. Reasons: What got you to $2-3M doesn’t scale past it. Not because effort drops. Because the system gets exposed: The offer still sounds good, but the economics aren’t tight. The ICP is defined, but not prioritized. Messaging gets responses, but not the right conversations. Outbound works, but only in bursts. Pipeline exists, but forecasting doesn’t. Each part works in isolation. The system doesn’t. Scaling past $10M requires alignment. 1/ Offer architecture If ROI isn’t obvious, velocity disappears. No amount of activity fixes weak economics. 2/ Market mapping Broad ICPs hide where revenue actually comes from. Growth accelerates when you concentrate on the segments that convert cleanly. 3/ Message precision Good messaging doesn’t persuade. It filters for buyers who already care. 4/ Outbound engine At scale, the pipeline can’t be accidental. It has to be created on purpose. If output depends on timing, referrals, or hero effort, it’s fragile. 5/ Revenue operations This is where scale either stabilizes or collapses. If data and handoffs aren’t tight, leadership is always reacting. 6/ Predictable scale Forecasting only works when inputs are controlled. When the system is stable, growth becomes boring. Miss one piece and everything downstream degrades. If outbound is part of how you plan to scale, it needs to be engineered with precision. Need help building a system that scales? Apply for our no-cost, no-risk pilot campaign. We’ll run a live campaign to show you how it should work before you make any commitments. Apply here: https://bit.ly/C17Pilot Follow Enzo Carasso 🧲 for outbound systems and growth. | 86 comments on LinkedIn

As Clay's largest user, we've learned a ton about scaling with the platform. Here are the 5 lessons I talked about on stage at Sculpt in San Francisco recently. (Link to video at the end). Clay Rules & Best Practices 1. Always Always Always Use Conditional Formulas on Paid Integrations. Seems super simple and obvious but getting your team in this habit helps to prevent unnecessary API calls and only trigger paid integrations when data truly needs enrichment. 2. Save Paid Data Externally Store enriched data in CRM or Supabase, not inside Clay only. I pay $30 a month for Supabase and just one of our databases amongst over 20 has 11.4M records. 3. Leverage Scheduled Sources Automate evergreen campaigns that continuously refresh and send new leads into workflows without manual input. This features is still under used by people that have cracked an outbound campaign for themselves and need to keep it on all the time without forgetting. 4. The formula generator can really do a lot. If you want the same thing to happen every time it sees an input, try this before using AI. Use formulas for: Cleaning and combining data If/Then logic Lead scoring State abbreviation Reminder: Clay formulas run on JavaScript, so almost any data transformation is possible. 5. The GEX AI Prompting Framework Prompting Rules The 10-Minute Manual Research Rule: Don't steal other people's playbooks. Think about when you are manually researching a company. What are you looking for if you had 10 minutes to research them? And how does that change your message? One Thing at a Time: Give AI a single, focused task, avoid multitasking prompts. Safeguards: Build filters for low-context outputs (e.g., return “purple” when uncertain). Use Examples: Always include examples in prompts for better accuracy. Meta Prompting: Ask ChatGPT “What questions do you have to improve this prompt?” before finalizing. Reasoning First: For high-stakes tasks (like industry classification), ask AI to output reasoning before its final answer. Handle Edge Cases: Be ready to prompt around specific constraints (e.g., tone, banned words, output format, length). Mindset Takeaway “AI is the smartest thing in the world built to solve Calculus problems. I promise it can write your emails, you just need to give it some TLC.” Overall: → Think like a system designer, not a button clicker. → Use Clay, Supabase, and AI together to build repeatable, data-rich workflows that scale outbound. https://lnkd.in/eMtqW3Ut