The Problem

Most small business owners know they should be doing something about their online presence. They've heard the word SEO. They may have even paid someone to "do SEO" at some point. But the reality of ongoing search optimization — monitoring rankings, analyzing competitors, identifying keyword opportunities, publishing fresh content, measuring what's working — is a continuous job. It's not a one-time fix.

For a business owner running a concrete leveling operation, or a dental practice, or a one-person consulting firm, that continuous job is essentially impossible to do well. There's no bandwidth for it. And hiring an agency to do it properly costs money that most small businesses don't have sitting around for something with a slow feedback loop.

The result: most small business websites just sit there. They were built once, maybe updated a few years ago, and they compete badly against businesses that are actively managing their presence.

What I Built

The system is a three-layer platform: a knowledge operating system (Kronos v2) that stores entities, relationships, and time-series data; an automation engine (n8n) that runs scheduled data collection and AI content pipelines; and an AI agent system (SpaceBot) that handles long-form content creation and code deployment. All three run in Docker Compose behind Traefik with automatic TLS.

Automated Data Collection

Ten automated pipelines run on fixed schedules without any human involvement:

Rank tracking — twice-weekly keyword position snapshots via DataForSEO, capturing position, ranking URL, SERP features, and search intent classification across all target keywords. Every snapshot is stored as both current state and historical trend.

Search Console analysis — weekly per-query impressions, clicks, CTR, and position from Google Search Console. The system automatically identifies striking distance keywords (positions 8-20 with high impressions) and significant position movers.

Traffic and conversion reporting — weekly GA4 data by landing page and channel, isolating organic traffic and calculating which pages actually drive business — not just rankings.

Technical health audits — weekly crawls of up to 50 pages via PageSpeed Insights, capturing Core Web Vitals (LCP, FCP, CLS, TBT), accessibility scores, and missing SEO elements. Issues are categorized by severity and tracked over time.

Competitor keyword discovery — monthly analysis via DataForSEO's domain intersection API, finding keywords where competitors rank in the top 20 but the client ranks below 50 or is absent entirely. New opportunities are automatically added to the keyword tracking universe.

AI Content Generation

Google Business Profile posts — three AI-generated posts per week, each calibrated with the client's brand voice, recent post history (for deduplication), review velocity, keyword targets, and service areas. A topic rotation algorithm cycles through seasonal relevance, review highlights, neighborhood spotlights, and service deep-dives to prevent repetition.

Review responses — every new Google review triggers an instant AI-drafted response via webhook. The system adjusts tone by sentiment: empathetic acknowledgment and offline resolution offers for negative reviews, specific detail callbacks and natural keyword weaving for positive ones. Every response doubles as a local SEO signal.

Long-form content — when keyword gap analysis identifies a content opportunity, SpaceBot's multi-process agent architecture generates full articles, deploys them via git, and measures the impact.

Human-in-the-Loop Approval

Nothing publishes without approval. Every piece of AI-generated content — GBP posts, review responses, content recommendations — is delivered as an email preview with HMAC-signed links. The business owner taps approve or reject. No login required. No dashboard to check. No app to install.

The approval system is configurable per client: data ingestion and observations auto-approve (zero friction for data flow), while content and strategy recommendations require human sign-off. Every approval, rejection, and outcome is recorded in a complete audit trail.

Closed-Loop Measurement

Every recommendation that gets approved and executed is tracked back to its outcome. Did the new content actually improve rankings? Did the keyword targeting change move the needle on traffic? The system measures before and after, tying specific actions to specific results. This is what separates the platform from a set of automations — it learns what works for each client in their specific market.

How It Started

My brother owns a concrete leveling business in Minneapolis. He mentioned he wasn't getting enough business from his website.

"Imagine if everything you needed to do to grow your online presence was nothing."

That conversation was the catalyst. I started building a system that would handle everything — monitoring, analysis, content, measurement — so the business owner's only job was approving what it produced. What began as a project for one client grew into a multi-tenant platform running across multiple businesses.

Why This Site Is Also a Test Case

The Angstrom Systems website is running on the same platform. This is intentional. The best proof that an autonomous SEO system works is applying it to the business that built it. If it can move the needle for a small AI consulting firm in a noisy market, it can move the needle for a concrete leveling business or a dental practice.

It also means I'm continuously improving the system against my own results — which creates a fast feedback loop that benefits every client running on it.

Where It Stands

The platform is operational with 10 active automated pipelines, a full management dashboard, and clients running on it. The system collects data twice weekly, generates and publishes AI content three times per week, responds to reviews in real-time, and delivers weekly performance summaries — all without the business owner lifting a finger beyond one-click approvals.

It's now a standalone product called Surfaced — also available on the products page. The multi-tenant architecture means adding a new client is configuration, not code — and the template system is designed to extend beyond SEO into other verticals.

Surfaced