AI Agent Workflow for Creating Pitch Emails

This one’s for Founders & Sales folks. I built an AI agent that cut my sales follow-up time by 90%. Not kidding. From 30 minutes per email... to 2 minutes. And I actually enjoy it now. Let me back up. I hate writing sales follow-ups. → Re-reading call notes → Trying to remember context → Spending hours wordsmithing Even with my system of organized ChatGPT folders with custom deal context, it still took forever. So I did what any founder would do. I built a tool. It sounds much harder than it actually was. I hadn’t built an AI agent before and it only took me 2 hours end to end. Here’s what I used and how it works. ⚙️ Built with: Relay.app (shoutout to Jacob Bank - love what you’re building!) Step 1: I trigger Relay to follow up with a particular deal in Hubspot. Step 2: Relay retrieves deal context from Hubspot (it’s made me much more diligent about making sure my data is up-to-date here) Step 3: Agent reviews the deal and decides if a follow-up is needed. It gives me the following output: Is a follow up required? Yes / No response What kind of follow-up is required? General check-in email, breakup email, nudge with resources (I provided these options for it to choose from). Why did it make this decision? This is really helpful because it gets me up to speed on the deal quickly—when did we last check in, what were their objections or concerns, when is the next expected touch point, and so on. Step 4: I approve or tweak. I tell the agent if it’s right or wrong, or provide context it may not have. Step 5: AI writes a draft email. The first draft hits me within ~20 seconds. I give high-level feedback (e.g., “focus more on timeline urgency”) if necessary. Step 6: AI revises the draft based on by input. At this stage I have an almost perfect draft. I make minor edits if at all and hit send. The whole process takes 2–3 minutes max. Are we all getting replaced by AI in 2 years? Probably. But for now, I’ve outsourced an annoying part of sales and it's amazing.

A no-brainer use of AI in sales is building GPT personas to write the perfect cold email. Here's exactly how I (and other top performing SDRs I know) do it step-by-step: ~~ THE OLD WAY You'd first get in the mind of your prospect. Write something. Tweak it. Come back with fresh eyes. Tweak again. Start to feel like this email sounds like spam. Finally send it. Over time, your intuition for "good" gets better. But, it's still time consuming. ~~ THE NEW WAY Use AI personas to "get into the mind of a prospect" and give line-by-line feedback. - Get more specific feedback. - Build an intuition on "good" quicker. - Enjoy a more energizing process writing with a soundboard. HOW I DO IT 1. Build your GPT persona "You are a sales leader for an enterprise sales motion. I need your help building persona guides so that my reps can reach out to people in their language. Give me a breakdown on [position]. Tell me their scope, personality type, goals, pain points, value Origami provides, and messaging themes/tone." 2. Give it context / mentality "Read this as if you are a skeptical SDR manager and you received a cold email. You are open to new ideas (if useful). But you hate getting pitched and will immediately delete an email if it feels salesy. Give me feedback on how it feels as you read it, line-by-line. Use the persona we built to guide you." 3. Get help editing "Also I want your help with edits. Do not re-write it for me. But provide specific feedback in lines where I don't follow these rules: - The first line must be personal, relevant observation - Tie it into the offer with relevant social proof - End with a soft CTA, offering something of value - Copy should be so specific to this target market / person that it couldn't be sent to another person and still make sense (s/o Nick Abraham). Ex: Reference their exact tech stack. Call out their specific compliance requirements. Use the metrics they actually care about. Mention the tools they use daily. Etc. - Be very concise. Do not use any extra words. - Focus on the prospect's pain points, not our solution." 4. Iterate until it's good to send ~~ GPT is a great collaboration partner. Saves a ton of time and energy. And helps w/ new verticals. But, it's not the be all end all. SOME DRAWBACKS: - It's still not there with one-shot writing the message. (though I imagine with enough context + training data on positive replies, AI could do a great job) - It will give you a lot of slop. So you should still be ruthless in what feedback you take vs not. Your taste / intuition matters. - In new markets, AI might hallucinate the persona info (pain points, goals, etc). Take its thoughts with a grain of salt.

Here’s an agentic workflow we built for one of our clients, and I wanted to give you a quick sneak peek under the hood 👀. The company was sending 50,000+ emails per month, but their sales process was manual, expensive (lots of SDRs), and inefficient. Their goal?💡 Automate SDR outreach using AI to reduce costs and scale faster. 🔴 The Problem: - SDRs Are Overloaded & Inefficient - The company had a large SDR team spending too much time on: --- Prospecting & research – Manually gathering lead data. --- Email & LinkedIn outreach – SDRs sending thousands of messages. --- Follow-ups & scheduling – Tracking responses, booking meetings. --- CRM updates – Manually logging all interactions. They wanted to automate this entire process, reduce headcount, and eliminate the need for external agencies. 🟡 Traditional SDR Workflow (Before AI) 1) Prospecting – Using Lucia & Outreach to find contacts. 2) Manual Research – SDRs scrape LinkedIn, Crunchbase, Google. 3) Email & LinkedIn Outreach – Sent via Gmail, HubSpot, Outreach. 4) Follow-ups – SDRs track responses and send reminders. 5) Meeting Scheduling – Calendly & HubSpot for booking. 6) CRM Updates – SDRs manually log everything into Salesforce. This works—but it’s slow, expensive, and doesn't scale well. 🟢 AI-Powered SDR Workflow (Agentic Model) We redesigned the workflow using AI Agents, cutting manual effort by 80% and enabling personalized outreach at scale. ✅ Prospect Research Agent – Scrapes lead data from Perplexity, LinkedIn, Crunchbase. ✅ Enrichment Agent – Uses Clay & GPT-40 Mini to enhance lead insights. ✅ AI Outreach Agent – Automates personalized emails & LinkedIn messages. ✅ Follow-up Agent – Auto-sends responses based on lead interactions. ✅ AI Scheduler Agent – Integrates with Calendly, HubSpot, cal.com. ✅ CRM Sync Agent – Pushes everything to Salesforce & HubSpot. This fully automated system replaces manual SDR work, reducing costs while increasing efficiency & scale. ⚙️ The Tech Stack 🔹 LLM: GPT-4o Mini (Optimized for speed & cost) 🔹 Agents: Prospect Research, Enrichment, Outreach, Follow-up, Scheduling, CRM Sync 🔹 Tools: Outreach, Lucia, Clay, HubSpot, Salesforce, Calendly/cal.com 🔹 Vector Database: Weaviate / Quadrant for knowledge storage & retrieval Looking to automate workflows with agents? Come talk to us at Lyzr AI #Agents