Latest / Sales Leadership with Fexingo: Quota Carriers, Sales Managers, and Revenue Teams / How Sales Reps Can Use Generative AI to Craft Better Emails
Transcript
- Lucas: So here's a number I haven't been able to shake: one mid-market SaaS company recently used GPT-4 to generate and A/B test five thousand different subject lines for a single campaign. They saw a 34 percent increase in open rates compared to their control group. That's not a lab experiment — that's a real pipeline lift. Luna: Five thousand subject lines? That would take a human writer weeks. What was the campaign about? Lucas: It was a standard outbound sequence targeting VPs of Operations in mid-size manufacturing firms. The product was a supply chain visibility tool. What's interesting is that the AI didn't just churn out variations of the same line — it suggested entirely different angles: urgency, curiosity, social proof, even humor. The winning subject line was something like 'Your competitors already know their inventory levels.' Luna: Right, that's specific and a little provocative. But I'm wondering — did they just let the AI run wild, or did they have guardrails? Lucas: Great question. The team set parameters: keep it under fifty characters, avoid trigger words like 'free' or 'guaranteed', and always include a personalized token like the company name. The AI generated batches of fifty, a human copywriter reviewed and selected the top ten for each segment, and then the A/B test ran against their usual template. So it's a human-in-the-loop system, not fully automated. Luna: That guardrail about trigger words is smart. Too many spam filters would flag those. So the 34 percent lift — did it translate to more meetings booked? Lucas: That's the key follow-up. Open rates went up, but the reply rate only improved by about 12 percent. So the subject line got them in the door, but the body still had to do the heavy lifting. The team actually used AI for the body drafts too, but they rewrote every single one before sending. The AI gave them a skeleton, and the rep added the specific insight or reference from the prospect's recent activity. Luna: So the real win might be efficiency, not magic. They saved time on drafting and could send more personalized emails per day. Lucas: Exactly. The rep who ran that campaign told me she used to spend about twenty minutes on a first-touch email — researching the prospect, writing, tweaking. With AI, she got that down to about six minutes. She could send ten more personalized emails per day. Over a month, that's an extra 200 touches. And touches lead to conversations. Luna: I want to get into the pitfalls, because I've seen some terrible ai generated emails. Ones that are clearly written by a machine — long, generic, no personality. How do you avoid that? Lucas: The biggest mistake I see is reps using the first draft straight out of the tool. The language is often too formal — words like 'leverage', 'optimize', 'utilize'. Real human beings don't talk that way. The best approach is to feed the AI your own best-performing emails. Give it five examples of your past emails that got replies, and ask it to generate variations in your voice. It's not a replacement for your style; it's a force multiplier for your best ideas. Luna: So you're saying the training data matters. If you feed it your own high-performing copy, the output will sound more like you. Lucas: Exactly. And another trap: over-personalization. Some reps try to cram in every detail they found on LinkedIn — 'I saw you commented on a post about blockchain in supply chain, and also that you were promoted last month, and your company just opened a new facility in Ohio.' It sounds creepy, not helpful. The AI can help you pick the one most relevant detail and weave it in naturally. Luna: That's a good rule of thumb: one personalization point per email, and make it relevant to the business problem you're solving. Lucas: Right. And this is where sales managers can really add value. They should be coaching reps on how to prompt the AI effectively. For example, instead of saying 'Write a cold email to a VP of Ops about supply chain visibility,' a better prompt would be 'Write a cold email to a VP of Ops in manufacturing. The prospect's company uses legacy ERP systems. Our product integrates with their existing stack and reduces manual inventory checks by 40 percent. Keep the tone direct and avoid jargon.' The more context you give, the better the output. Luna: That makes sense. It's like giving a junior rep a brief instead of just telling them to make a call. The prompt is the brief. Lucas: Yeah. And I think this is where the show's listeners might find a practical takeaway: if you're a sales rep, experiment with one email sequence this week. Take your best-performing email from the last month, feed it to ChatGPT or whatever tool you use, and ask for five new versions targeted at different buyer personas. Then test them against your control. You might be surprised. Luna: We've been talking about outbound, but what about follow-ups? AI can help there too, right? Lucas: Absolutely. Follow-ups are actually where AI shines, because they're formulaic by nature. You can ask the AI to generate a sequence of five follow-ups with increasing urgency, each one referencing a different value proposition. One rep I know uses a prompt like 'Write a follow-up email for a prospect who opened the first email but didn't reply. Mention the ROI case study from the healthcare vertical. Keep it under 100 words.' She says her reply rate on follow-ups doubled. Luna: That's a great specific tactic. I also want to flag a risk: if every rep in your company uses the same AI tool with similar prompts, your emails could start sounding the same. Differentiation becomes harder. Lucas: That's a real concern. We're already seeing it in some industries. A few weeks ago, I got three cold emails from different SaaS vendors, and they all had the same sentence structure — 'I noticed that is doing, and I thought you might be interested in.' It was clearly from the same template. The solution is to customize your prompts and inject your own voice. Also, use the AI for inspiration, not final copy. Edit heavily. Luna: If these conversations have moved your work forward in some small way, a couple of dollars a month is genuinely what keeps them going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's listener support that lets us keep this ad-free and focused on what actually works. Appreciate anyone who chips in. Luna: So back to the practical side — what metrics should a sales manager track to see if AI email tools are actually helping? Lucas: I'd look at three things: first, the number of personalized emails sent per rep per day. If that goes up without a drop in reply rate, you're winning on efficiency. Second, the reply rate itself — compare it month over month. If it stays flat or goes up while volume increases, that's a clear win. Third, the time saved — ask reps to self-report how much time they save per email. That translates to more time for high-value activities like discovery calls or demos. Luna: What about the quality of replies? Not just quantity. Lucas: Good point. You want to track the percentage of replies that lead to a meeting booked. If that goes up, the emails are not just getting opened — they're compelling enough to start a conversation. In the case I mentioned earlier, the meeting booking rate improved by about 8 percent. Not huge, but meaningful when you're scaling volume. Luna: So the overall picture is that generative AI is a tool for sales reps, not a replacement. It helps with volume and speed, but the human touch — context, judgment, relationship-building — remains critical. Lucas: Exactly. And I think that's the right frame. The reps who will thrive are the ones who learn to use AI as a co-pilot, not a crutch. They still need to understand the buyer's business, ask good discovery questions, and build trust. The AI just handles the repetitive parts. Luna: One more thing: any advice for sales leaders on implementing AI tools across a team without causing disruption? Lucas: Start small. Pick two or three reps who are open to experimenting, give them access to a tool, and have them document what works. Then share those wins with the broader team. Don't mandate it from the top down — that breeds resistance. Let the early adopters become champions. Also, set clear guidelines about data privacy: no feeding customer names or proprietary info into public AI models. Luna: Good point. That's a compliance issue that can't be ignored. Lucas: Yeah, and it's an area where legal and sales need to align. Some companies have built their own internal AI tools that run on secure servers. That's the gold standard. But even with off-the-shelf tools, you can anonymize data. Use 'Company A' instead of the real name. The AI doesn't need to know who the prospect is to write a good email. Luna: Alright, anything else you'd add for a rep listening right now who wants to try this tomorrow? Lucas: Yes. Pick one email you already sent that got a reply. Paste it into an AI tool and ask for five variations with different emotional hooks — one curiosity, one social proof, one fear of missing out, one direct benefit, one question. Send each to a small segment of your list. Track which gets the best response. Use that as your new control. Rinse and repeat. That's it. That's the playbook. Luna: I like it. Simple, measurable, and it puts the rep in control of the tool rather than the other way around. Lucas: Exactly. And if you do that for a month, you'll have a data set that tells you exactly what your specific buyers respond to. That's worth more than any generic template.