Keeping your outreach human when everyone is using AI

Contributors

1. Pranjal Kukreja, CEO at Optima Bags

   Pranjal Kukreja | CEO | Optima Bags The concrete tactic: I write the opener of every outreach piece manually, even when AI drafts the rest. The first two sentences are original, reference something specific and recent about the recipient's work, and could not have been generated by any template tool. Everything after that can be structured with AI help 6 but the hook that earns the open has to be human-written and specific. The workflow: before drafting any pitch or cold email, I spend 3-5 minutes reading actual recent output from the person or publication I'm reaching out to. One specific detail from their recent work 6 a product they covered, a story angle, an opinion they expressed 6 goes into the first sentence. That specificity immediately signals that this isn't a batch send. Where leaning on AI backfired: I ran a test where AI drafted the full pitch for a batch of 25 outreach emails to travel accessory journalists and editors. Response rate was around 2.8%. I then rewrote only the openers for the same list 6 everything else stayed identical 6 and resent. Response rate jumped to 19%. The AI body was fine. The AI opener was the problem. Journalists scan for signals of mass personalization within the first line, and AI openers all share a recognizable rhythm that signals templated outreach immediately. The single lesson: AI is excellent at structure and saves significant time. It's poor at demonstrating genuine attention, which is the only real currency in outreach. 6 Pranjal Kukreja, CEO, Optima Bags

2. Raj Baruah, Co Founder at VoiceAIWrapper

   AI can draft, but a human has to own the first line and the ask. At VoiceAIWrapper we send a lot of cold outreach to agency owners, and the version that consistently lands is one where AI gives me a rough draft and I rewrite the opener and the request by hand. The opener has to reference something only that specific person would recognize. A recent post, a vertical they work in, a tool they mentioned. That is the part a template cannot fake, and it is the part that decides whether the email gets read. Three tactics carry most of the result. First, very short subject lines, often one word, because they read like a note from a person and not a campaign. Second, a first line written for that one reader, never a swappable variable. Third, I reply personally to every human reply, even a one-line one. The reply signal alone keeps more of my mail out of spam than any deliverability trick. The backfire taught me the rule. Early on I let AI write a full sequence, polished end to end. It read perfectly, and that was the problem. Perfect reads as mass-produced. Open rates were fine, replies were near zero, and a chunk got filed as promotions. The fix was not better writing. It was making the first sentence specific enough that no one could mistake it for a blast. Polish is not the goal. Recognition is. People reply when the first line proves you actually looked at them, and no model can manufacture that for you.

3. Hans Graubard, COO & Cofounder at Happy V

   In our category, AI drafts fail in a way most operators don't catch: they sound credible but quietly slip in structure-function claims we can't legally make. "Restores vaginal flora," "treats BV," "balances hormones" -- language an OB/GYN on our advisory board would flag instantly, and language an LLM produces by default because it's pattern-matching wellness copy on the open web. So every journalist pitch, retailer deck, and partner email gets human review before send. AI drafts the skeleton; I or our marketing lead rewrite against the same claims matrix we use for on-label copy. Slower, but the alternative is a buyer or reporter passing without telling you why. The backfire that taught us: an AI-assisted partner email referenced a condition by name in a way that read as a treatment claim. No reply. The wins come from short notes that reference something real -- a study, a customer pattern, a formulation choice -- and stay inside what the label says. Credible in regulated health isn't warmth. It's accuracy.

4. Ben Frederick MD, Founder at Dr. Frederick's Original

   My team sends a lot of outreach. We had AI draft a batch that went out to a few dozen retail buyers, and the copy was polished, warm, even funny. Every reply we got back was some version of "thanks, not interested". When I pulled up the thread, I could not find a single line that referenced the recipient's store, their product mix, or why we would be a fit for their customers specifically. So now AI builds the skeleton, and every email includes at least one detail that shows someone spent five minutes looking at the recipient's world, like a mention of something they recently stocked or a comment about their customer base. The AI saves us hours every week on structure and formatting, and each email gets a few minutes of human research on top of that.

5. Hugo Gomez, CEO at Abogados NOW

   My team sends outreach to Spanish-speaking audiences who can spot a generic template faster than most people expect. A poorly translated AI draft sounds robotic and signals that whoever sent it does not speak the language or understand the community. We lost a media placement last year because an AI-drafted pitch used phrasing that was grammatically correct but culturally off. The editor told us it read like it was written by someone who had never had a real conversation in Spanish. After that, I changed my workflow. AI can pull together background research, find the right contact, and draft a skeleton. But every outreach piece gets a final pass where someone on my team rewrites the opening and closing lines by hand, in their own words, referencing something specific about the recipient's recent work. That manual layer adds about four minutes per email. It also brought my reply rates back to where they were before we started leaning on AI drafts.

6. Jason Levin, CEO/Founder at Memelord.com

   We run a marketing team of 4 at Memelord.com and send outreach at real volume. AI drafts the first pass on almost everything now, but the tell is always the same: generic openers that could've been sent to literally anyone. Our rule is nothing goes out until it references something specific about the person, a recent post, a launch, a joke that actually landed. That one line takes 30 seconds to add and it's the entire difference between a reply and the trash folder. We got burned once letting a template go out with the wrong company name still in it because someone trusted the AI draft and skipped the human read. Lost the placement and felt dumb for a week. Now every pitch gets eyes on it before it leaves the building, no exceptions. AI writes fast, but speed without a real read just turns your outreach into spam with better grammar.

7. shankar@getboomerang.ai, Founder @ Boomerang at Getboomerang

   I let AI draft, but I never let it source. The moment a pitch leans on generic insights, it reads generated and dies. What lands is anchoring every message on one thing only I could say, like the enterprise deal where three execs went dark and a warm intro to the VP of Sales Enablement, who was not even our buyer, unlocked the whole account. Where it backfired: I once let AI write a researched pitch end to end and a detection tool flagged it 100% AI. The tell was not grammar, it was the absence of a real story. Now AI handles structure and speed, but the specificity is always mine.

8. Amit Agrawal, Founder & COO at Developers.dev

   The moment AI takes over your decision making process within Outreach, you're no longer building relationships with individuals, rather you're now building a list of Spammers. As a digital growth expert with almost 20 years experience in outreach management, I've seen how AI has powerful potential as a research tool; however, it can also be dangerous to have it do the job of a human voice. The biggest mistake I witness with today's teams is letting the machine create the hook of the email. We have developed a human-in-the-loop process: we scrape and summarise prospects' recent activity, annual reports, and industry news using AI to identify real pain points for the prospect. This is how we achieve efficiencies through saving multiple hours of manual research. But I will never allow AI to write the final draft of the email. The final draft will always be written by someone who synthesises the research into one, specific and relevant question for the prospect. If the prospect believes they're receiving a templated email, they will stop reading it. I can remember a campaign where we automated our follow-up cadence too much. Our automated follow-up messages were gramatically perfect and technically accurate; however, they lacked social intelligence because the prospect was going through a major change in their publicly traded company. We lost that deal due to not communicating with the prospect at an appropriate level of context or emotional weight. This is the risk of using AI as an email communication tool; you risk losing the ability to understand the context within which you're trying to communicate with your audience. When you're looking for the needle in the haystack, use AI to help you find the needle but the final decisions regarding the best message to send should be made by the person that has the awareness of what a "magnet" is. Vit should be mentioned that Outreach should enhance productivity of your outreach, not hinder your ability to connect and communicate with your audience.