Summary of "How I Automated an SEO Agency with 15 AI Agents (No-Code)"

High-level summary

Ben (creator) demonstrates an automated end-to-end SEO content production system built for an SEO agency serving fashion e‑commerce brands. The system uses a 15‑agent, no‑code AI architecture (Relevance AI + Make.com) to handle the full workflow: research → content brief → draft → finalize → publish (to Webflow/WordPress/Shopify/Magento).

The primary business impact reported by the agency was a massive increase in content output (and lower per‑piece cost), correlated with meaningful organic click growth for several clients. Ben emphasizes this isn’t a silver bullet (seasonality and other factors apply), but volume was the biggest lever.

Key framework / architecture (playbook)

Multi-layered agent architecture (core pattern)

SEO Director agent: orchestration, CRM updates, final publish step.
Manager agents: supervise groups of subagents and evaluate outputs (examples: Research Manager, Content Brief Manager).
Subagents: task specialists (keyword research, topic research, competitor scraping, each brief type, draft writers, final writers, backlink inserter, product‑data retriever, image fetchers).

Content production workflow (mapped to agents)

Trigger (from CRM/Airtable) → SEO Director.
Research phase:
- Keyword agent (SEMrush).
- Topic research (web scraping, Google, Perplexity).
- Competitor research (scrape top pages).
Content brief creation (brief templates vary by page type).
Drafting (content drafter uses brand examples).
Finalization (type‑specific final writer; pulls product DB, images, internal links; optional backlink agent).
Publish (post to CMS via Make.com connectors) and update CRM.

Human-in-the-loop gates

Content briefs are returned to the SEO Director/agency for review and edits before final content is written/published.

Reliability controls

Manager agents evaluate subagents’ outputs and can send work back for rework (quality control loop).
The system supports retries on integration errors and includes monitoring for publish failures.

Concrete processes, templates and integrations

Trigger chain: Airtable trigger → Make.com webhook → Relevance AI agent orchestration.
- Airtable automation requires a Team plan to run scripts; Ben shows an example run‑script that posts the Airtable record ID to a Make.com webhook.
- Make.com flow: receive webhook → Get Record (Airtable) → HTTP module to call Relevance AI agent.
Content briefs: tailored per content type (blog, category, product description, product review, landing page). Each brief includes:
- Summary, target audience, keywords, article structure, key points.
- Internal/external link ideas and competitor analysis.
Product database & assets:
- Final writers pull product metadata (images, prices, links) via vector search over the product DB to include tables, product blocks, and internal linking automatically.
SEO and research tools:
- SEMrush for keyword/backlink research.
- Perplexity (or Google scraping) for topic research.
- Vector search for product matching.
- Stock image retrieval for missing product images.

Key metrics, KPIs and results mentioned

Client organic click growth (agency screenshots):
- Small client A: ~125 clicks/day → ~300 clicks/day over ~3 months.
- Small client B: ~125 clicks/day → ~300 clicks/day (similar trend).
- Larger client: ~750 clicks/day → ~1,900–2,000 clicks/day (steep increase after rollout).
Content throughput and cost:
- Prior manual process: ~150 pieces per 2.5 months (≈2/day) at ~$50 per piece.
- Automated system: up to ~150 pieces per day at ~$1.50 per piece.
- Implied change: ~75× daily output; cost per piece reduced from $50 → $1.50 (~97% decrease).

Caveat: seasonality and other marketing factors may also have contributed to traffic growth; the agency attributes the biggest impact to increased output volume.

Concrete examples / demo artifacts

Demo topics: “Best Nike running shoes” and other Nike examples for blog, product review, and category pages.
Outputs shown:
- Generated blog posts.
- Category descriptions with product blocks/tables.
- Stock/product images.
- In full client system: FAQs and backlink insertions.
Publishing examples posted to Ben’s Webflow demo site (layout/FAQ rendering limited by demo site structure).

Actionable recommendations (what to copy/adapt)

Use a multi‑layered agent design (director → managers → subagents) to model real human workflows and add quality gates.
Maintain content brief templates per page type; automate brief generation from research outputs.
Integrate a product database (vector search) to auto-populate product data, images, and internal links for e‑commerce content.
Use manager agents to evaluate and request rework from subagents — this is the key quality control mechanism.
Keep a human‑in‑the‑loop check on briefs or final outputs, at least initially, to maintain brand voice and compliance.
Include dedicated backlink and internal‑link agents if you use SEM tools (SEMrush) to automate link insertion.
Use Airtable as a client‑facing dashboard + Make.com webhooks to trigger agent orchestration; provide simple fields (URL, topic, content type, language, optional keywords/competitors, notes).
Budget consideration: SEMrush and Perplexity APIs add cost — test cheaper alternatives or manual keyword inputs where needed.

Operational & risk notes

Content quality and ranking: the presenter acknowledged controversy about AI content ranking and withheld full release until client results were verified. The system’s main effect appears to be volume rather than guaranteed per‑piece ranking improvements.
Integration fragility: publish steps can error — include retry/resilience and monitoring.
SEO controls: ensure briefs include competitor gaps/opportunities and conversion-optimizing elements (category descriptions are important for ranking and conversions).
Legal/brand risk: client names were withheld; always include review steps and brand voice examples.

Tech stack and tools used

Core agent / no‑code: Relevance AI (multi‑agent builder), Make.com (automations & integrations).
CRM / workflow: ClickUp (client used it); Ben demo used Airtable for triggers.
SEO / data tools: SEMrush (keyword + backlink research), Perplexity (topic research) or Google scraping.
CMS publishing connectors: Webflow, WordPress, Shopify, Magento (via Make.com).
Additional: vector DB for product search, stock image retrieval tools, Google Docs (optional).

Limits of the demo and product availability

Ben’s public demo omitted SEMrush/Perplexity integrations due to cost; some features (backlink agent, FAQ automated placement) were left out.
Ben is not releasing the full template for free but will share the Airtable webhook script and a Figma overview; consultations and deployments are available for more clients.

Presenters / sources

Presenter: Ben (creator; runs an AI automation community of ~3,000 members).
Client: unnamed SEO agency specializing in fashion e‑commerce brands (case studies/screenshots referenced).