This week in numbers (real, from our system)
- 🤖 AI agents running: 19
- 📝 Content published: 63 (blog RU 16, EN 22, Altezza 25)
- ⚙️ Generated programmatically: 4
- 📥 Leads in the system: 335 (+0 in the last 7 days)
Figures as of 2026-09-30 — computed by code from the DB and files, no manual entry.
Short answer: Building in public means sharing how we actually run things, including the failures. Our AI agents draft articles, build SEO pages, support sales and read analytics, while people set direction and approve what ships. Content automation works when humans stay in the loop. Exact figures live in a separate block.
Most companies show the polished result and hide the machinery. We'd rather show the machinery. This diary covers how our AI agents run a content factory, what works, and what keeps breaking.
Why are we building in public?
Look, we sell AI and automation to small and medium businesses. It would be odd to do that without using it on ourselves. Building in public keeps us honest. If an agent produces weak copy, you can see it. If a pipeline stalls, we write it down.
It also helps you. If you run a company or lead marketing, you get a realistic picture of content automation, not a demo.
How do our AI agents run the content factory?
Think of it as an assembly line. Each agent has a narrow job, a clear input and a clear output.
Writing articles
An agent takes a topic and keywords, then drafts a structured article. Rules are strict: no invented data, no fake testimonials, a fixed structure. A human reviews before anything is published.
Generating SEO pages
Another agent builds pages for specific queries and services. It's repetitive work, which is exactly why it suits automation. Quality checks catch thin or duplicated pages.
Supporting sales
Agents qualify incoming requests, prepare context for a conversation and draft follow-ups. The person still talks to the client and closes the deal.
Reading analytics
A reporting agent summarises what pages perform and what doesn't. It suggests the next topics. We decide which suggestions are worth acting on.
What broke this week?
Here's the thing: something always breaks. Typical failures are an agent ignoring a formatting rule, a draft that sounds generic, or a handoff between agents that silently drops context. We fix these with tighter prompts, validation steps and better logging, not with hope.
Where do people stay in charge?
People choose the strategy, the tone and the final publish decision. Agents remove routine: first drafts, formatting, data pulls. That frees the team for the work that needs judgement. The goal is to augment people, never to replace them.
Frequently Asked Questions
Does content automation mean no human editing?
No. Every piece goes through human review. Automation speeds up drafting, but responsibility for accuracy and tone stays with people.
Can a small business build something similar?
Yes, if you start small. Automate a single repetitive step, measure the result, then expand.
Why not publish the numbers in the text?
We insert verified figures as a separate block, so the narrative stays honest and the data stays accurate.
What is the biggest risk of AI agents in content?
Unchecked output. Without rules and review, agents produce confident but generic or wrong text.
Conclusion
Building in public is uncomfortable, but it makes the work better. If you want to see how AI agents could take routine off your marketing team, get in touch. We'll share what we've learned, including the mistakes.
Top comments (0)