Most AI consultants talk about AI. We run our own marketing with it — and we decided to show you how, instead of telling you.
The reason is simple. When you hire a firm to build AI agents for your business, you have every right to ask: "Do you actually use what you sell?" Our answer is yes — and this article is the proof. Drafted by an agent. Reviewed by a human. Published automatically. Exactly how we build our ads.
This isn't a buzzword — it's what marketing looks like in 2026
"Agentic marketing" has stopped being an experiment. According to McKinsey's 2026 research, autonomous AI agents can take on as much as two-thirds of routine marketing tasks — research, content generation, budget planning. The marketer's role shifts: from the person pushing the buttons to the person setting the direction.
That's our stance too. The agents do the repetitive work. People decide what makes sense. Here's what that looks like in practice — the four stages from the diagram above.
The four stages, concretely
01 · Research. Before we write a single line of ad copy, an agent reads what already works in the niche — formats, length, what actually makes a Bulgarian small-business owner stop scrolling. We use NotebookLM to synthesize that into a short brief with concrete recommendations, not generic platitudes.
02 · Creative. The agent writes the copy, the headlines and the visual direction — usually several variants at once. We don't bet on "one brilliant idea"; we bet on letting Meta test variants and tell us which one wins.
03 · Launch. Here most people picture someone clicking through Ads Manager by hand for half an hour. For us, the campaign, ad sets and the ad itself are created directly through the Meta Marketing API — programmatically, with the right pixel and targeting wired in from the first moment. The very evening we're writing these lines, we launched a restaurant ad set in exactly this way.
04 · Optimize. After launch, the numbers are in charge. What gets impressions, what drives clicks, which headline holds attention — data from Meta and from our own analytics (Humblytics) decides what stays active and what stops. No ego, no "but I personally prefer the second one."
What we deliberately do NOT automate
This is the important part — and the reason you shouldn't fear "a robot running your budget."
No ad and no article goes out before a human has approved it. The agent prepares; the human judges. The judgment of what we promise, which numbers are honest to claim, what tone to take with a Bulgarian customer — that stays with people. Because that's exactly where a mistake costs trust, and trust doesn't come back from an A/B test.
of our ads and articles pass a human review. The agent never publishes a brand-risk decision on its own.
Honest about the stage we're at
We won't show you a case study with an impressive ROAS and borrowed numbers. Our ad campaigns are young — deliberately, we start with small budgets and build slowly, because for a Bulgarian small business that's the sensible approach, not aggressively pouring money into a cold audience.
What we're showing you is the pipeline — real, working, transparent. The same system that runs pragma-lab's marketing is the system we build for our clients. The difference is that with you, you'll see every stage with your own eyes.
Why this matters for you
If you run an online store, a restaurant, a practice, or any small business in Bulgaria, the pain is the same: marketing eats time you don't have. A pipeline like this doesn't make you redundant — it frees you from the routine part and leaves the decisions that genuinely need a human with you.
It's not magic. It's engineering — clear processes, human review, no jargon. And it already works, every day, for us.