Hi all, What platforms do you recommend for creating outreach messages for campaigns? Ideally something that can automatically generate compelling copy using our ICP and USP details. Keeping msges human and overall value centric. Would love to hear what’s worked for you.
Open to custom GPT or Claude solutions aswell.
Claude is your best solution for everything messaging. Although it's important to create something like an OS built into Claude that will generate custom messages for every contact you feed it. However I usually advise using this for initial messages not the whole campaign sequence. The initial messages should achieve or have one goal! To get a response. After a response a human should get in the loop. ( The best outcomes I have gotten from follow up message created by Claude was with loads of context) This works great if you are doing a fair Volume. If volume is Much then you'll have to create a more complex AI system to handle it giving that human feel.
I’ve been using Clay plus Claude for this. You can feed it your ICP and USP, combine that with company specific context, and use Claude to generate personalised outreach. Next to it, build some deterministic rules (skills) in Claude for tone, key words, text size per outreach channel, and focus points
+1 to clay + claude. one thing i'd add, have someone skim a sample before each batch goes out. tone is usually fine, its the wrong facts about the prospect that slip through
We've Outboundry.com with an AI-powered campaign generation feature that lets you create personalized outreach copy using your preferred AI model. Simply define your campaign goal, tone, language, and context, and Outboundry generates tailored messaging based on each lead’s LinkedIn profile. You can also configure fallback channels across Email, LinkedIn, and LinkedIn InMail to keep your outreach flexible and effective. This is working for our users.
I actually find Claude and chat to be really generic Rakin not sure if you are experiencing this. Yet to find a very good solution but for now I’ve been hashing multiple times a good enough template and then getting broader pinpoint variables researched and pulled through. It's more work but you can actually get replies
Rakin, I find Claude to be really good for copywriting as long as you give it enough context about your outreach goal, ICP, tone, offer, and likely some good cold email copywriting frameworks like AIDA (Attention-Interest-Desire-Action), PAS (Problem-Agitation-Solution) etc
Christian S. Tim M. Kshitij B. Kedar K. Fiona W. Natasha O. I’m building a Claude skill/project for writing cold emails that the whole team can use. Would any of you like to collaborate on it? How it works right now - you fill in a USP and ICP questionnaire, and Claude uses that to write the email. The problem: even with good inputs, the emails come out generic. They’re accurate, but bland, with no personality and no angle that makes someone want to reply. What I think will fix it - giving Claude a library of really strong cold email examples to learn from.
happy to help Rakin. examples help, but in my experience the generic feel usually comes from two things:
no real angle per prospect. give it one specific observation per lead (a post, a hire, a launch) and make it build the email around that one thing, not the whole icp
the same ai phrases leaking in. "hope this finds you well", "i noticed that", "leverage", "seamless". a banned phrase list in the skill helps a lot
i've got a list of the ai-sounding lines my pre-send check flags, happy to share it for your skill. dm me
Happy to help Rakin! One thing I’ve noticed when working with Clay + Claude is that the emails often become generic before Claude even starts writing. If you ask it to find an angle and write the email in one go, it tends to fall back on the ICP and produce something that’s accurate but not particularly interesting. So what has worked better for me is splitting those steps. First, focus on finding a genuine reason to reach out now. Keep the source and date attached. If there’s no solid reason, don’t force an email. Then use that context to write the message. Ideally, each type of signal should lead to a different angle or opening question, rather than just changing a few words in the same template. And yes, on the example library, I’d include good and bad examples, with a short explanation of why one works better than the other. Otherwise, Claude can end up copying the style of the examples without really understanding what makes them effective. I’d also track how often the team actually sends the emails without having to rewrite them. That’s a pretty useful indicator of whether the output is getting better. Happy to share more of what I’ve been experimenting with and collaborate on this. Good luck anyways 🙂
Rakin, for the team skill, add a small test set alongside the example library. Use the same 10 prospect briefs after each change: a few strong signals, a few weak ones, and at least one with no usable evidence. Keep these separate from the examples the model sees. Have it return the source-backed observation, a buyer-problem hypothesis, the email, and a send/hold decision. Score whether every factual claim is supported, whether the offer follows from the observation, and whether the question is easy to answer. A new hire supports ‘you’re hiring’; it doesn’t establish that their pipeline is broken. That gives you a way to test whether a new example actually improves the skill rather than just changing its voice. In an agentic outbound workflow, keep this check separate from sending and stop follow-ups when a reply arrives. Our full ten-case checklist and a worked example: autonomousgtm.ai/blog/test-cold-email-skill-before-sending
