Why you're getting this: Why you're getting this: You've interacted with Haven or AdTribe before, and this is my weekly-ish update on my journey to building an AI-native agency. Zero pressure to stick around, just click unsubscribe if you don't want to get it.
Alright, let’s get into it.
Town hall with my team
I asked my Claude agent to count how many times I said "AI" in Slack this year. The answer was 114.
And yet if you asked any of my media buyers what "AI-native" actually means for their day-to-day, most couldn't tell you. That's on me.
I've been saying the word constantly without actually explaining what I mean. For a lot of the team, that probably just creates anxiety. People hear "AI" from their CEO over and over and start wondering what it means for their job.
I've let that sit unaddressed for too long. So I'm running a townhall.
The main thing I want to do is explain where we're heading as a company, and what skills we need to acquire altogether to win.
Using ChatGPT to draft an email is not AI-native. That's just using a tool. What I actually mean is something closer to this:
Old model: a media buyer does the work manually and delivers value to the client.
New model: the buyer provides the expertise, the judgment, the context and AI handles the execution. The buyer doesn't go anywhere. But what they spend their time on changes completely.
What I'm going to ask from each person is simple to say and harder to do: document every decision you make inside the accounts. Not just what you did, but why.
That context is what makes the system smart over time. Without it we're just running AI on top of a mess.
I also want to be honest about what's coming. It's a lot of work. New tools, new habits, thinking differently about what good work actually looks like. Some people will find it energizing. Some will find it uncomfortable. I'm not going to pretend it's easy.
But the upside is real.
The buyers who actually push through this and build the skill of working alongside AI are going to be significantly more valuable. Not just at AdTribe, but anywhere in this industry.
The gap between people who figure this out and people who don't is going to get very wide, very fast. I want everyone on our team on the right side of it.
From meme to roadmap
A few weeks ago I started meeting with developers to figure out what becoming AI-native actually requires. It went exactly like this:
Humbling.
Turns out wanting AI and being ready for AI are completely different things. Our data was scattered across five tools, naming conventions were inconsistent, half our processes weren't documented anywhere.
You can't build something smart on top of that.
So we slowed down. Week after week of conversations, each one surfacing new problems and new frameworks, until it stopped looking like a meme and started looking like a plan.
We landed on three phases:
Phase 1: Foundation
One database that pulls everything in: ad performance, Slack history, call transcripts, creative files, account context.
Becoming an AI company means becoming a data company first. Until the data pipeline is clean and connected, building agents on top of it is pointless. Garbage in, garbage out. Everything else waits until this is done.
Phase 2: The agents
Eight of them, built in order of biggest time save. Starting with account monitoring (replacing a 30-60 minute daily manual checklist) and reports.
Our fastest buyer does reports in 12 minutes. Our slowest takes 58. Same accounts, same output, 4x gap.
Then working up to the communication helper, since some buyers spend over 3 hours a day just on client Slack replies.
Phase 3: Feedback loops
Every time a buyer corrects the AI output, that correction gets logged against the account context.
After 12 months we'll have a system trained on 8 years of our media buying history plus thousands of real corrections from the team. No other agency will have that. That's the moat.
Target: cut fulfilment cost from 50% to 25% of revenue within the year.
Lot of work ahead. But for the first time it actually looks like a plan.
Talk soon,
Dmitry