A couple of weeks ago I sat down with Mike Kolev, CEO and co-owner of Carbon Box Media.
His agency is doing $356k a month, 70% gross profit with 17–23% month-on-month. And he hasn’t lost a single team member as they transitioning to become an AI-native agency.
We asked him how they:
Use AI where it actually matters (80% of efforts do not impact the revenue)
Operating through hypergrowth
Acquisition systems: volume game, vendors, ads, outreach
Where agencies should start with AI
Future of agencies + adoption rate discussion
Watch the full podcast (links at the bottom).
The 80% rule nobody wants to hear.
Mike's single most important point on AI isn't about which tool to use.
It's this:
"70–80% of the things that you're going to introduce as changes won't have any kind of direct impact on the revenue."
Most founders are running around implementing AI everywhere and “feeling” productive. Mike's point is that most of it is noise. The 20% that moves the needle is almost always in delivery… the part that costs the most, takes the longest, and keeps founders trapped in operations.
So that's where he started.
His first AI build: a tool for the SEO division that could deliver 90 blogs in 90 days, previously done by $23,000/month in Upwork writers.
With the tool, delivery dropped from 60–70 days to hours. He can now onboard unlimited clients for that service.
The second: "Advertisable".
It’s an AI creative tool for paid media (70% AI, 30% human). Why? Because every paid agency in the world suffers with creatives. They built it internally first, then started packaging it as a product.
The AI example every agency should steal.
They had a trading client who'd been in the market for 6–7 years. At some point he couldn't differentiate himself from newer competitors.
Mike's team also built a sentiment analysis tool.
They scraped comments from the client's competitors' YouTube videos. Categorised comments as neutral, negative, positive. Found the five topics the ICP kept coming back to.
Those five topics became a lead magnet. The paid team ran ads against them. Result: record sales in six to seven years.
"It's one workflow. 10 steps. Go to YouTube, scrape. That's it."
Where to start if you have zero AI right now.
Mike's framework is straightforward.
Step one: download the brain.
The biggest problem in most $1-5M/year agencies is that the smartest thinking only exists in the founder's head. The strategist's head. Senior leadership's head. Nobody else can access it, so every decision escalates up.
Build a knowledge base trained on how the key people think. Now anybody in the agency can tap into that thinking without needing a meeting.
Step two: Fix acquisition follow-up.
Many agencies are slow to respond to leads. An AI system acting as a virtual SDR nurturing, following up, sequencing, etc., captures what would otherwise fall through.
"If you're a minus-one professional and you add AI, you multiply by 10. That's still minus. If you're a one, you become a ten."
Full conversation is linked below.
Before you watch it… what's the one thing in your agency that only exists in your head right now? The thing that creates bottlenecks every single week because nobody else can answer without you?
Hit reply and tell me and I’ll get back with a personalised answer to what you should build first.
– Romans