What if AI ran your entire marketing department?

Employee like working agents, not a dashboard

Okeanos needed Dashboards too, they help visualize the business. But the real problem wasn't visibility. It was capacity. Two people can't run marketing, sales, follow-ups, content, ads, SEO, and analytics WHILE overseeing actual installations for a company trying to 4x its installs.

So the question became: what if the system didn't just show you what's happening but actually did things for you?

15 agents, each with one job.

That's what I built. One agent watches ad performance and reallocates budget when cost-per-lead spikes. One writes social posts matched to what's working this week. One scores incoming leads based on their behavior on the site. One sends the first follow-up email within minutes, not days.

They don't operate alone. There's a hierarchy. The CMO agent acts like a real Chief Marketing Officer and reviews, delegates, overseas the work of other agents. It also sets the strategy, weekly plan, updates or checks with the humans through Telegram as the main conversational interface along with a Web UI that included dashboards and chatbots.

Think of it less like "AI magic" and more like a very organized marketing department that never sleeps.

How much is automated?

Almost everything.

Lead comes in → scored and routed. Automatically.
Follow-up email → sent within minutes. Automatically.
Ad budget shifting because one campaign's tanking → done. Automatically.

Two people don't have time to approve every email. That was the whole point.

The exception: anything high-risk.

For example, someone leaves a bad Google review? The agent drafts a response. A human reads it before it goes live. You don't let AI handle someone's frustration unsupervised.

But, we didn't place top 3. And, I think about this a lot.

Some teams that placed higher built systems with a toggle. Automatic or manual. Let the business choose what to hand over.

We chose for them. We looked at two people with no time and said: you need full autonomy. Here's 15 agents. Go.

And honestly? That confidence might have been the problem.

We learned the CGS Advisors' Agentic Enterprise Maturity Framework in class. The core idea: it's not just about how smart your agents are. It's whether the organization is ready for them. There are layers to this: strategic mindset, technical maturity, operational readiness and they need to move together.

Okeanos wasn't starting from nothing. They were already running a 6-agent system with memory and real-time call analysis. Not beginners.

But our proposal jumped them from "6 agents with guardrails" to "15 autonomous agents running the entire marketing function." That's not the next step. That's three steps…

The pitch didn't land either

We led with what excited us, the architecture. 15 agents, Telegram chatbot, coordination layers, API integrations, autonomous workflows. Technically impressive.

But Okeanos isn't evaluating a tech stack. They're evaluating whether this actually helps their company.

If I pitched it again, I'd use the continuous customer journey framework from class because it has two sides that mirror each other.

The customer side: a homeowner moves from "I want a pool" to searching online, to requesting a quote, to comparing options, to buying, to owning and eventually needing service. Each step shapes the next.

The firm side: that same journey maps to marketing, then sales, then service. What Okeanos does at each stage determines whether the customer moves forward or disappears.

I'd walk through both sides together. Here's what the customer experiences. Here's what Okeanos has to do. Here's where agent augments or offloads.

Framed as an experience, not an architecture.

The real takeaway

This project taught me to build agents. But the more valuable lesson had nothing to do with the tech.

CGS Advisors' Beyond the Bot paper talks about how companies don't need an AI strategy but they need an intelligence strategy. Instead of asking "where can we use AI?", start asking: what's the right mix of human and machine intelligence for this business? Which tasks need human judgment? Which ones don't? How do they work together?

I skipped that question. I went straight to exciting part of building.

The agents work. I'm proud of the system.

But if I started this project again, I wouldn't open Cursor first.

I'd map the customer journey (both sides) and figure out where the real gaps are. I'd look at the maturity framework and be honest about what Okeanos is actually ready for, not what's technically possible. I'd consider the intelligence architecture, the right mix of human and machine at every stage based on their strategy and vision, before writing a single line of code.

And then I'd build agents. Probably fewer. Definitely with a dial instead of a switch. And I'd pitch the experience, not the architecture.

The number of agents doesn't matter. What matters is building the right system, not the most impressive one.