Quick Summary
- Client
- Philippine Herbal Laboratories, Visayas (B2B)
- Problem
- Posting without knowing what content actually worked
- Solution
- Apify · OpenAI · Google Sheets · Apps Script · Zapier
- Result
- Weekly AI-generated content insights delivered automatically every Monday
- Timeline
- 3 weeks — September to October 2025
A Visayas-based Philippine herbal brand with a B2B-focused Facebook presence was posting regularly but had no structured way to evaluate whether their content was reaching and resonating with their Ideal Customer Profile.
The Challenge: Data Without Insight
The brand wasn't struggling to post. They were posting consistently. The problem was they couldn't answer three critical questions about their content:
Is this content speaking to our actual ICP? Their target audience is B2B — manufacturers, researchers, institutional buyers. Were their posts positioned for that audience, or were they accidentally attracting the wrong people?
Are we hitting engagement benchmarks? Likes and shares are visible, but engagement rate relative to follower count and industry benchmarks tells a very different story.
What should we post next? Without a systematic way to evaluate what worked, every content decision was being made from intuition rather than evidence.
Manual audits were theoretically possible but not practical. A proper content review takes hours — scraping posts, calculating engagement rates, comparing against benchmarks, evaluating ICP alignment, and writing up recommendations. Nobody had time to do this weekly. And doing it monthly meant acting on data that was already too old to be useful.
The real problem: it wasn't a content strategy problem. It was a data infrastructure problem. The brand needed a system that could do the analysis automatically — so the team could spend their time acting on insights, not generating them.
The Solution: A Multi-Layered AI Agent System
I built a fully automated social media intelligence system — two AI agents working in sequence, each with a distinct role, scheduled to run weekly without any manual trigger. By Monday morning, the team has a complete analysis of the previous week's content waiting for them.
Data Collection — Apify Facebook Scraper
The system starts by scraping the brand's Facebook page using custom Apify actors — pulling post captions, engagement metrics (likes, shares, comments), media links, and images. Custom actors were built to capture multiple images per post simultaneously, giving the AI richer visual context for analysis than a standard single-image scrape. All data flows directly into Google Sheets as structured rows — no manual export, no copy-pasting.
Engagement Scoring — Google Apps Script
Once the data is in Google Sheets, a custom Apps Script calculates engagement rates for each post and scores them against predefined industry benchmarks. Posts that exceed benchmarks are flagged as high performers. Posts that fall below are flagged for review. This replaces manual calculation — and because it's formula-driven, it updates consistently every week without drift or human error. Image processing is also handled here via Apps Script rather than through Zapier, reducing automation task usage and keeping costs lower.
Agent 1: ICP Alignment Analysis (OpenAI)
The first AI agent reviews each post through the lens of the brand's Ideal Customer Profile. It's trained on the brand's voice, positioning, and target audience — so it evaluates posts the way a strategist embedded in the business would. For each post, it answers: 'Is this speaking to the right person? Does the messaging, visual, and angle align with what a manufacturing partner or research institution would respond to?' This produces an ICP alignment score and a brief qualitative note per post.
Agent 2: Weekly Content Strategy Summary (OpenAI)
The second agent zooms out. It takes the full week's post data — engagement scores, ICP alignment notes, image analyses, and benchmark comparisons — and generates a human-style strategic summary. This isn't a data dump. It reads like a weekly briefing from a content strategist: what worked, what didn't, what patterns are emerging, and what to focus on in the coming week. Specific, actionable, written in plain language. This second workflow runs 6 days after the first — staggered deliberately to prevent overlapping executions and ensure the summary draws on a complete week of data.
Delivery — Google Sheets + Zapier Orchestration
Everything is delivered back to a master Google Sheet — structured, side-by-side, easy to scan. The team sees post content next to engagement score, next to ICP alignment rating, next to the AI's strategic note. The weekly summary sits at the top. Zapier orchestrates the entire sequence — triggering Apify, moving data between platforms, scheduling the two OpenAI agent runs — all hands-free.
The Stack
Apify
Data collection layer, Facebook scraping with custom multi-image actors
Google Sheets
Central data layer, structured storage, engagement scoring, side-by-side output
Google Apps Script
Computation layer, engagement rate calculation, benchmark scoring, image processing
OpenAI
AI agent layer, ICP alignment analysis and weekly strategic summary generation
Zapier
Orchestration layer, schedules, triggers, and connects all platforms in sequence
Visit Zapier
Google Drive
Image storage, post visuals saved and referenced for AI analysis
What the Team Has Now
Every Monday morning, without anyone pressing a button, the marketing team opens their Google Sheet and finds a complete content intelligence report for the previous week.
- Every post scored against engagement benchmarks — high performers flagged, underperformers identified, patterns visible at a glance
- ICP alignment evaluated per post — the team now knows whether their content is actually reaching and resonating with their B2B target audience, not just generating generic engagement
- A strategic weekly summary they can act on immediately — not raw data, but clear recommendations for what to do differently in the coming week
- Zero manual effort — no scraping, no spreadsheet formulas, no report writing. The system runs every week whether or not anyone thinks to trigger it
What Makes This Different From Standard Social Media Analytics
Most social media analytics tools give you data. This system gives you interpretation.
- ICP-trained analysis, not generic engagement metrics — the AI knows who this brand is trying to reach and evaluates content against that specific standard, not a one-size-fits-all benchmark
- Multi-image analysis — the system reads and interprets multiple images per post, not just text captions. Visual content is evaluated as part of the ICP alignment assessment
- Two agents, two perspectives — the post-level agent and the weekly summary agent serve different analytical functions. One is granular, one is strategic. Together they give the team both the detail and the big picture
- Fully automated and self-sustaining — no dashboard to log into, no report to request. The intelligence arrives on schedule, week after week, without anyone managing the process
Is This Right for Your Business?
This type of AI agent system makes sense when you have an active social media presence but no structured way to evaluate whether it's working for your specific audience; your marketing team is making content decisions based on intuition rather than consistent data; you have a clearly defined ICP that you're trying to reach — and you want to know whether your content is actually positioned for them; or you want strategic content insights delivered automatically — without adding workload to your marketing team.
The system can be adapted for Instagram, LinkedIn, or other platforms depending on data availability. The core architecture — scrape, score, analyse with AI, summarise strategically — applies across channels.
Running on manual workarounds?
One discovery call and you'll know what's worth automating in your operation, what isn't, and what a working system would look like.