This Thursday, June 4th, 9 AM PDT I'm sitting down with Jason Smith, CEO of Klue, to talk about what it really takes to "re-found" an established company in the AI era. Jason completely redesigned his company and how work gets done with an agent-first mindset. Headcount changed–a lot–and why the hardest part isn't the tech, it's the culture. Don't miss it. Register here!

And now back to today’s post….

For most of the past decade, "personalization at scale" meant putting someone's first name in an email subject line and calling it dynamic content. It was mail merge dressed up in SaaS language–or maybe just mass mail merge not gussied up at all. Nobody was fooled, least of all the buyers on the receiving end. It also wasn’t that hard to do anyway.

The gap between what personalization at scale claimed to be and what it actually delivered was one of the defining disappointments of the martech era. The technology promised individualized experiences. Optimizely at one point raised a monster round on the promise of personalizing the internet–every website custom to whoever the specific visitor was.

Not creepy at all. Also, like 3D TVs, not a trend that lasted.

The reality most of us settled on as acceptable was slightly-less-generic batch emails, sent to segments of ten thousand people and called "personalized" because everyone in the segment shared one or two characteristics.

That gap is genuinely closing. Not because the aspiration changed, but because the data infrastructure and the AI tooling have finally caught up to it. But you can still trigger that creepy factor if you’re not careful.

Here's what's actually working.

Start with your data, not your templates

Most personalization efforts fail before they start because the underlying data is wrong. CRM fields are incomplete. Firmographic data is 18 months stale. Product usage signals live in a data warehouse that's disconnected from your marketing stack. You can have the best personalization engine in the world and it will produce garbage if it's working from garbage inputs.

The prerequisite work is having good data. For big companies this means a data audit. Map what you actually know about each contact and account at the point of outreach versus what you wish you knew. The gap is usually embarrassingly large, and it's the gap that your personalization is working around rather than working with.

For smaller companies without legacy data, this can mean using better sources. We’re testing or have tested over a dozen at Skyp–if you’re interested in this subject, reply. Maybe I’ll spill some candy.

Close the most important gaps first: job title verification and seniority (LinkedIn data decays fast and most CRMs are worse), company funding and growth stage (a Series A company and a Series C company are different buyers), technology stack (what they're currently using tells you what the switching cost is and what the integration story needs to be), and behavioral signals from your own product if you're PLG.

Third-party intent data is valuable but overrated by most of the vendors selling it. Instead use Exa or another real-time check so that when you do reach out, you’re using the most up to date information. If you could be wrong, skip it.

Why most outbound still feels like spam

The bar for outbound has raised permanently, and it has raised because AI-generated personalization has made the bad version ubiquitous, and the templated stuff just looks lazy. Buyers can spot both instantly.

I got one in LinkedIn this Sunday. “Circling back” used to be human, now it is 100% coopted by AI. And nothing hits less hard than a generic conversation with “another founder”.

Bilal, if you’re reading this, reply–maybe we can get this sorted for you.

It reads as noise. More importantly, it damages your relationship with the specific buyers you need to reach over the next 12 to 18 months, before they've ever given you a real chance. As people wise up to this they (like me) are unfriending these new “friends” on LinkedIn. No space in our lives for laziness, or at least laziness without some thoughtful AI effort.

Outbound personalization that works requires a genuine insight about the recipient's specific situation at this specific moment. That means identifying what's actually changed or is changing for them: a new VP of Sales who inherited a broken outbound motion and is hiring a GTM Engineer, a company that just expanded into a new market where your product solves a different set of problems, a competitor that just raised prices and left their install base looking for alternatives.

AI can surface these signals and help draft the message. But the insight has to be real. If you can't articulate specifically why you're reaching out to this person right now, your outbound isn't personalized.

Personalizing the wrong thing

Most growth teams spend 80% of their personalization energy on acquisition and 20% on lifecycle. That ratio is usually backwards.

Once someone is a customer, you have first-party behavioral data that makes genuine personalization tractable in a way that acquisition personalization almost never is. You know what features they use and which they ignore. You know where they get stuck in onboarding. You know how their usage compares to similar customers who expanded versus those who churned. You know when they last logged in and what they did when they were there.

That data can drive personalized onboarding flows that respond to actual behavior rather than time elapsed. It can trigger expansion conversations at the moment when usage signals that a customer has hit the ceiling of their current tier. It can identify at-risk accounts weeks before the renewal conversation, when there's still time to change the outcome.

The revenue impact of personalized lifecycle programs is almost always higher than the revenue impact of equivalently-resourced acquisition personalization programs. The data is better. The conversion rates are higher. The cost per dollar of incremental revenue is lower.

If you're a growth leader with limited resources and you have to choose where to build personalization infrastructure first, build it in lifecycle. That's where the math is most in your favor. Of course, eventually you’ll need top of funnel also–but you’ve got to start somewhere.

The Goldilocks of personalization

Begin with the ICP and timing questions: why is this the right person or company to reach out to, and why now? Make sure your data is current–that you’re not asking about people or activities that happened four or six months ago. Even being a week late for key moments can be far too late.

Choose where to start–existing relationships or new ones. If you’re using excellent data, and are thoughtful about your outreach, it’s fine to use AI to scale up upsells, cross sells and recovery campaigns. But stick with small batches, and be careful. Don’t just point a new fancy AI tool at your entire CRM and click “send it”.

For new relationships, there’s a lot of data out there. We get it for customers, and there are customers who get it for themselves. Exa is a great resource–doing live web searches in a very targeted way ensures you’re using the latest information, though not everything is available on the web.

Just don’t go too deep. We tried the “I noticed you hired Bob to run your sales team last week”–trust me, it did not go well. Too much personalization is 100% a bad thing. That works if you actually know Bob–if you just found it on the internet, it’s creepy.

An ask

Julian and I are looking to talk to more teams about their GTM. Not just outreach, but everything. If you want to grab some 1:1 time, we’re happy to share what’s working for us that might not make it into the newsletter, and listen to what you’re working on and what might be slowing you down. If you’ve read this far… just reply, or grab a time.

And, if you didn’t know, Skyp is built on the premise that personalization starts with signal, not templates. When your outreach is triggered by something real and specific happening in a prospect's world, the message hits differently. That's the distinction that separates the outbound that gets responses from the outbound that gets ignored. If you want to see it in action, book a demo. We can also talk GTM more broadly, if you like.