What's behind the automation service we sell? This page shows the architecture of the AI-powered lead-research pipeline we built for our own growth — no marketing gloss, just the real system.
Finding a business, reviewing its website, noting what's broken and writing an email specific to it — serious time for even one prospect. For hundreds, it's unsustainable. So we built a system that automates the process end to end.
Local businesses are scanned via the Google Places API with city × sector rotation; SMB filters applied.
Each business's website is fetched: PageSpeed measurement, screenshot, contact-email discovery.
An LLM with vision reviews the site using structured output, and records only verified findings.
Each business gets a personalized email draft grounded only in step 3's verified findings — the door to invented claims stays shut.
Nothing sends automatically: mandatory human approval, then export. Automation produces; the decision stays human.
The pipeline's sibling: a service that watches the inbox over IMAP, applies deadline guards, and pushes AI reply drafts to Telegram for one-tap human approval — verified by 37 automated tests.
Honest note: This isn't a client project; it's internal tooling we built for our own studio, still waiting for its production window — which is why you won't see made-up "results" numbers on this page. What we're showing is the architecture: the automations we build for your business are engineered with the same principles and the same care.
Tell us — we'll be honest about what should be automated, and what shouldn't.