Outbound
I build outbound systems end to end: sending infrastructure, enrichment, intent detection, segmentation, reply handling. Then I operate them, so they don't decay the way most outbound does.
Fifteen Years Of This
In-house
Previously Head of Outbound at Comp AI. Designed and operated the complete outbound engine that powered growth across the business.
Client builds
Outbound built for SaaS, ecommerce, fintech and more. The whole pipeline, end to end, adapted to how each industry actually buys.
My own companies
I run this on my own pipeline, so I find the cracks before clients do.
01
Nobody reads your copy if it lands in spam. I build separate sending infrastructure for every domain, warm it properly before it touches a prospect, and keep monitoring placement week over week, not open rate, because open rate tells you nothing about whether you're in the inbox or the spam folder. The system decides if the words matter. Get the system wrong and the best copy in the world never gets read.
02
Nobody reads a cold email. They read the first line, decide if you're worth thirty more seconds, then either keep going or delete it. Writing that line once is easy. Writing a version of it for every person on a list of ten thousand is an engineering problem, not a writing problem. I build the pipelines that pull the right detail for each contact and generate copy that reads like it was written for one person, because it was. Then I keep rewriting it as the reply data comes back.
03
People don't reply because you asked nicely. They reply because something in the email made them think, this is actually about me. That's not a trick, and it doesn't scale by hiring more writers. It scales by building enrichment that surfaces the one detail about a company or a person that makes the email true only for them. The psychology is real. The system is what lets you apply it to five hundred people instead of five.
04
Most outbound targets a title. I build the infrastructure that targets a signal, enrichment pulling in real-time data, not a static list, so the email goes out the moment someone actually has the problem, not whenever the campaign happens to run. A list of fifty thousand people with the right title is worse than five hundred people my systems just flagged as dealing with the exact thing you solve.
This is the same problem I solved with Alertedly, just pointed at a different question. Tracking companies for competitive signals and tracking companies for buying signals are the same system. Intent isn't something you guess at. It's something you build infrastructure to detect, and then keep tuned as the signals move.
05
The angle changes depending on who's reading. What gets a marketing lead to reply isn't what gets an operations lead to reply, even if you're selling them the exact same thing. I build the segmentation and routing logic that decides which hook a contact gets before a single email goes out, so the system picks the angle, not a merge field with a name pasted in. When a segment stops replying, I change the angle rather than sending more of it.
Judgement
Most of a good outbound system is fixed logic that just needs to run reliably. AI only comes in where judgment is the actual bottleneck, is this a real signal or noise, does this reply need a human or can it stay on autopilot. Everything that reaches a prospect is evaluated before it goes live. Anything that sends externally keeps a review gate until it's earned the right to remove one.
Case Study
Global specialty ingredients supplier
Their list was the problem, not their pitch. A CRM built over years, full of people who had changed jobs, addresses that bounced, and companies nobody had checked on since the record was created. Every send was burning domain reputation on contacts who no longer existed.
So the fix started with infrastructure. Separate sending domains, warmed properly before a single prospect saw them, with placement monitored instead of open rate. Then enrichment: every dead contact traced to their replacement at the same company, every account re-verified, and the sector each one operated in pulled in as a field the copy could actually use.
That last part is what made the copy work. An ingredients buyer in beverage cares about different things than one in personal care, so the system routes the angle by sector before the email is written, not after. Replies come back into a classifier that sorts interested from out-of-office from not-now, and the interested ones land with their sales rep with the context attached.
Domains get watched, dead records get replaced as they go stale, and the copy gets rewritten when a segment's reply rate starts sliding. Their reps stopped opening spreadsheets entirely.
The Five, Applied
8.4%
Reply rate after rebuild
3,610
Contacts replaced
Contact
If you're hiring for GTM or outbound, this is what I'd own from day one. If you want it built for your business, tell me who you're targeting and I'll reply within 24 hours with where it's leaking and what I'd fix first.