AI made automation cheap. It didn’t make it smart out of the box though.
I’ve been caught in two different client feedback loops recently where they’re trying to identify critical oversight in some of their processes. It will surprise no one that both of these led us right back to AI as the culprit.
Most teams are automating everything right now: outbound sequences, account scoring, content, nurture flows. And most of them have no idea if any of it works, because they never proved it manually first.
I’ve had principles I’ll live and die by for probably 7 years now but candidly haven’t revisited or updated them in a while. This week though I felt compelled to add a new one.
Here are the original principles. 👇
My newest principle: Earn the right to automate.
Pre-AI, your intelligence lived in your head, in your coworker’s head. It lived in notebooks, on whiteboards, in random Notes on your phone. All filled with insight about what the market wants, how your team really ships products/campaigns/you name it, what’s working / what’s not working, what bets you’re taking next.
In this “everybody must use AI” era, we skipped right over context and jumped straight to agentic everything.
Earning the right to automate means going back to that phase on purpose. Not indefinitely, but long enough to actually understand what you’re automating.
If the underlying process is broken, automation just accelerates the failure. Bad messaging doesn’t improve at scale, a scoring model built on flawed assumptions doesn’t get smarter at volume — it prioritizes the wrong accounts faster than any person can catch (until you’re digging deep when you’ve missed pipeline goal by 3x).
I’ve watched this play out twice now. Three months in to the “AI everything” era, pipeline is down. The team can’t diagnose it because the feedback loops are buried inside a system someone vibe coded so no one really understands it. The automation ran so fast it outpaced their ability to learn from it. And now oops, the team missed their goals.
The manual phase is the research phase. Remember when we used to do all research manually? 🙃 Just my two sense but if you don’t have a contextual layer embedded in your AI processes, you are just operationalizing a guess. Scale this up to a Series B SaaS company and now you’ve also got 250 teammates operationalizing their guesses. Sheesh.
Kieran Flanagan wrote about this this week from an architectural angle — he calls it the Context layer, the intelligence foundation that every agent in your system has to read from. His framing is org-wide which is more than true. I’m only credible enough to speak on demand gen. And truly this “manual phase” is how you build the demand gen piece of this contextual layer. You build this contextual layer by doing the work, understanding what’s working, building a POV for yourself. Then you can encode that understanding into your systems.
For outbound, that means writing and sending messages yourself before you build the sequence. For scoring, it means reviewing accounts by hand until the real intent patterns become obvious. For nurture, it means talking to prospects at each stage before you write the copy. Ads same thing. And so on.
Automation rewards the teams who understand what they were automating.
See ya next week,
Kaylee ✌