Technology Stack
Modern AI &
Engineering Stack
InsightRx is built with the same tools used by production AI engineering teams — not academic toy frameworks.
OpenAI API
GPT-4o for agent reasoning & synthesis
AI Core
Claude
Strategy & recommendation agent reasoning
AI Core
LangGraph
Agent-to-agent orchestration framework
Orchestration
Python
Core pipeline, API clients, data processing
Infrastructure
Streamlit
Interactive dashboard & demo interface
Interface
Figma Make
UI prototyping & system design
Design
Claude Code
AI-assisted development & debugging
Dev Tools
GitHub
Version control & project management
Infrastructure
pandas
Data wrangling, normalization & transformation
Data Layer
Codex CLI
OpenAI terminal agent for code generation
Dev Tools
Why InsightRx Matters
The Search Landscape Is
Changing Faster Than Most Realize
Three converging shifts are rewriting the rules of marketing intelligence. InsightRx is built for all three simultaneously.
01 / SEARCH
Traditional SEO Is Now Table Stakes
Google rankings alone are no longer enough. AI Overviews, zero-click search, and answer engines are reshaping discovery — requiring brands to optimize beyond traditional SEO.
02 / AI SEARCH
AEO and GEO Are the New Frontier
AI-driven platforms like ChatGPT and Google AI increasingly control discovery. Brands must now optimize for machine readability, authority, and answer visibility.
03 / AGENTS
Agentic AI Is Transforming Marketing Work
AI's biggest marketing impact isn't faster content — it's automating analysis, surfacing strategic opportunities, and accelerating decision-making.
The Problem
The Future of Marketing Intelligence
Is Agentic.
InsightRx uses AI-driven workflows and specialized agents to transform fragmented marketing data into strategic insights and operational efficiency.
Time Lost to Manual Analysis
Marketing teams waste hours exporting, cleaning, and stitching data before strategy can even begin.
40%+
of marketing time spent on data prep, not strategy
Siloed, Disconnected Data
SEO, social, paid, and website data remain fragmented—hiding cross-channel growth opportunities.
4+
disconnected platforms the average marketing team manually bridges
Stale Data, Late Decisions
By the time reports are reviewed, strategic decisions are often based on outdated performance.
3–7
days average lag between data and strategic action
Dashboards That Report, Not Recommend
Traditional analytics tools show performance but fail to prioritize actions or next steps.
0
traditional BI tools that tell you what to do next
Data Layer
Four Live Data Sources.
One Unified Intelligence.
InsightRx connects to live marketing APIs — not static exports. Every agent is grounded in real-time data from the platforms that define modern marketing performance.
Web Analytics
Google Analytics 4
The Analytics Agent ingests sessions, users, conversion events, traffic sources, engagement rate, and behavioral funnel data — detecting anomalies and surfacing performance shifts in real time.
sessionsconversionsengagement_ratetraffic_sourceevent_count
Search Intelligence
Google Search Console
The SEO Agent pulls keyword-level impressions, click-through rates, average position, and page-level performance — identifying ranking opportunities, content gaps, and technical indexation issues.
queriesimpressionsclicksctrposition
Competitive SEO
Search Intelligence
Provides competitive keyword rankings, backlink intelligence, domain authority metrics, and content gap analysis — giving the SEO Agent a complete picture of where organic opportunities exist relative to competitors.
keyword_rankingsdomain_scorebacklinksgap_analysis
Social & Paid Media
Meta Graph API
The Social Agent consumes ad performance, audience insights, post engagement, reach, and creative-level metrics from Facebook and Instagram — identifying what's working and where budget is being wasted.
impressionsreachctrroasaudience_insights
⚙️ Pipeline Workflow
System Architecture
How InsightRx
Actually Works
A six-stage pipeline from live data ingestion to ranked strategic output — designed to replace the manual analysis workflow entirely.
📡
01
API Grounding
Live data pulled from GA4, GSC, search intelligence & Meta
🤖
02
Agent Spawn
Six specialized agents initialized with domain context
🔍
03
Parallel Analysis
Agents analyze independently: SEO, AEO/GEO, social, analytics
🔄
04
A2A Handoffs
Agents pass structured findings to each other via LangGraph
🧠
05
Strategy Synthesis
Cross-channel patterns scored & prioritized
🎯
06
Action Output
Ranked recommendations with rationale & impact scores
Agent-to-Agent Communication
Unlike standalone AI tools that operate in isolation, InsightRx uses LangGraph to orchestrate structured handoffs between agents. Each agent produces a machine-readable output that downstream agents consume as input context.
AnalyticsAgent anomaly signals shape SEOAgent priority queue
AEO/GEO gaps inform StrategyAgent's content recommendations
SocialAgent engagement data calibrates RecoAgent's channel weighting
Workflow Transformation Impact
InsightRx maps the existing marketing analysis workflow, identifies every manual step, and replaces it with an automated agent sequence. The transformation is measurable.
Manual report pull: 3–5 hours → Agent pipeline: ~90 seconds
Cross-channel synthesis: hours → instantaneous A2A handoffs
Strategic output: static chart → ranked, rationale-backed action plan
🤖 AI Agents
Agent Architecture
Six Specialized Agents.
One Coordinated System.
Each agent has a defined role, a defined data source, and a defined output schema. They don't just run in sequence — they communicate. This is what makes InsightRx an intelligence system, not a pipeline of isolated AI calls.
Google Analytics 4
● Live
Google Search Console
● Live
Search Intelligence
● Live
Analytics Agent
DATA LAYER · GA4 API
ga4_api
anomaly_detection
conversion_analysis
→ JSON signal
SEO Agent
SEARCH INTELLIGENCE
gsc_api
ranking_api
gap_analysis
→ JSON signal
AEO / GEO Agent
AI SEARCH READINESS
entity_audit
schema_analysis
ai_visibility
→ JSON signal
⬢ A2A Handoffs · LangGraph Orchestration
Strategy Agent
CROSS-CHANNEL SYNTHESIS · LANGGRAPH A2A
Receives structured JSON from all four ingestion agents via LangGraph handoffs. Identifies cross-channel correlations invisible from any single-platform view, scores opportunities by strategic priority, and passes a ranked action plan to the Recommendation Agent.
a2a_input
priority_scoring
pattern_recognition
langgraph
Recommendation Agent
ACTION OUTPUT LAYER · STREAMLIT RENDER
1
Optimize title tags on 14 high-impression, low-CTR pages — estimated +18% CTR improvement
SEO↑ High
2
Implement FAQ schema on top 8 pages — brand absent from AI Overviews for 23 target queries
AEO↑ High
3
Shift 22% of Meta budget to retargeting — ROAS 3.1× vs 1.4× differential detected
Paid↑ Med
📊 Recommendations Engine
Output Layer
The Recommendations
Engine
InsightRx doesn't produce reports. It produces a ranked, rationale-backed action plan — ready to execute, not interpret.
Agent Outputs Collected
All 5 upstream agents submit structured JSON findings
→
Opportunity Scoring
Each signal scored by impact, effort, and strategic alignment
→
Ranked Action Plan
Prioritized recommendations with rationale and channel
Sample Output — Ranked Recommendations
1
Optimize title tags on 14 high-impression, low-CTR pages — estimated +18% click-through improvement based on GSC data patterns
SEO
↑ High
2
Implement FAQ schema markup on top 8 product pages — brand currently absent from AI Overview responses for 23 target queries
AEO
↑ High
3
Shift 22% of Meta ad budget from broad audiences to retargeting — ROAS 3.1x vs. 1.4x differential identified by Social Agent
Paid
↑ Med
4
Publish 3 topical cluster articles to build entity authority for generative AI visibility — 0% GEO presence currently detected for primary category
GEO
↑ Strategic
5
Fix 3 crawlability issues blocking 12% of content from Google indexing — technical debt flag surfaced by SEO Agent audit
Technical
↑ Med
Search Intelligence Layer
SEO. AEO. GEO.
Three Layers. One System.
Modern search visibility has three distinct layers. Most brands optimize only for one. InsightRx analyzes all three simultaneously — identifying exactly where a brand surfaces, where it disappears, and what to do about it.
🔍 SEO
Search Engine Optimization
The foundational layer — ranking in Google and Bing for high-intent keywords through technical health, content authority, and strategic link architecture. InsightRx's SEO Agent covers this comprehensively using live GSC and search ranking data.
Real-time keyword ranking gap analysis vs. competitors
Technical health: Core Web Vitals, crawlability, schema coverage
Content opportunity scoring from live GSC impression data
Internal linking and on-page optimization recommendations
🤖 AEO
Answer Engine Optimization
The AI assistant layer — structuring content so ChatGPT, Perplexity, and Google AI Overviews extract and cite your brand as the authoritative answer. This requires a fundamentally different optimization approach than traditional SEO.
Identifying high-intent queries that AI assistants answer directly
Content structure for machine extractability: Q&A, schema, FAQs
Auditing whether brand content is cited in AI-generated answers
Topical authority clusters that LLMs recognize and trust
✦ GEO
Generative Engine Optimization
The frontier layer — ensuring brands surface when AI engines proactively recommend products and services in conversational responses, without a user clicking a link. This is where M2M brand evaluation begins.
Brand presence audit in generative AI responses across platforms
Machine-legible brand signals: entity definitions, structured data
Positioning content to influence AI retrieval and training patterns
Share-of-voice monitoring in zero-click AI recommendations
SEO Layer
Human searchers clicking blue links
Google · Bing · SERPs
AEO Layer
AI assistants citing brand as the answer
ChatGPT · Perplexity · AI Overviews
GEO Layer
Agents recommending brands without search
All Generative AI · M2M Evaluation
🔗 Grounding APIs
Grounding Layer
Live API Connections
Google Analytics 4 API
Sessions · Events · Conversions · Audiences
Google Search Console API
Queries · Impressions · CTR · Position
Search Intelligence API
Rankings · Competitor Gaps · SERP Features
Meta Graph API
Ad Performance · Audience · Engagement
AI Answer Monitoring
Perplexity · AI Overview citation tracking
Live SERP Snapshots
Real-time competitive SERP positioning
Why Grounding APIs
Change Everything
Change Everything
Traditional marketing AI tools analyze exports — last week's CSV, last month's report. Grounding APIs change the equation: InsightRx agents query live data sources directly, so every recommendation reflects current conditions, not historical snapshots.
Always-Current Intelligence
Because agents query live APIs, every insight reflects what's happening right now. A traffic anomaly that started yesterday surfaces in today's recommendation — not in next week's report review.
Cross-Source Correlation
When agents combine live GA4 data with live GSC rankings and live Meta performance, they identify correlations that static, siloed reporting can never surface. The pattern between a ranking drop and a traffic anomaly becomes visible in real time.
Foundation for M2M Intelligence
Grounding APIs are the infrastructure that will ultimately power machine-to-machine brand evaluation — where AI agents query brand data directly to make recommendations on behalf of users, without any human search behavior involved.
🔄 A2A / M2M Communication
Forward Direction
How AI Agents
Evaluate Brands
Tomorrow's customers won't always be humans. AI agents are increasingly evaluating, comparing, and recommending brands autonomously.
Traditional Search
User
has a question
↓
Google Search
keyword matching
↓
10 Blue Links
user chooses
↓
Human Decision
manual evaluation
Agentic Search
User
has a question
↓
AI Assistant
interprets intent
↓
Agent Network
queries brand data
↓
Brand Evaluation
authority · structure · clarity
↓
Recommendation
no link required
InsightRx helps brands become discoverable inside AI-driven recommendation systems — auditing machine-readability, authority signals, and entity clarity before competitors do.
A2A / M2M flow
User Query
natural language
→
AI Assistant
interprets intent
→
Recommendation Agent
queries brand layer
→
Brand Evaluation
authority signals
→
Brand Authority
structure · schema · E-E-A-T
→
Recommended Brand
zero-click
What this means
AI Agents
AI systems increasingly evaluate brands without traditional search behavior.
M2M Validation
AI systems verify information through other AI systems and trusted data sources.
Zero-Click Discovery
Recommendations happen before users ever visit a website.
Brand Readiness
InsightRx measures machine-readable authority and visibility.
Live Product Preview
The Agent Pipeline in Action
Each tab walks through one stage of the InsightRx multi-agent workflow — from raw data intake to a ranked, ready-to-execute action plan.
app.insightrx.ai / pipeline
Data Intake
47,210 pts
✓
→
Insight Agent
12 insights
✓
→
Strategy Agent
5 opps ranked
✓
→
Execution Agent
5 action briefs
✓
→
Evaluation Agent
Report ready
✓
Opportunity Score
5.37
↑ +0.4 vs last run
Top Opportunity
CTR Optimization
Botox cluster · Pos 4.1
Traffic Trend
+12%
vs prior 4-week avg
Engagement Trend
+8%
Session depth improving
Actions Ready
5
4 high priority
Agent Status
5 / 5
● All Complete
Critical: Engagement–Conversion GapCritical
4m 48s avg session but only 1.2% conversion — 3.1× below industry median. Strong intent, broken funnel.
↗ Execution Agent brief: Move CTA above fold, simplify booking form to 3 fields.
High: CTR Gap on Top Keyword ClusterHigh
12,400 monthly impressions at Pos 4.1 yielding 1.9% CTR vs 6.2% expected. ~520 clicks/month unrealized.
↗ Execution Agent brief: Rewrite title tags with price signal and local modifier.
⬡ Ask the AI Agent● Live
"What should we prioritize this month?"
Based on your data, the top 3 actions are: (1) Rewrite title tags to recover ~520 clicks/mo · (2) Move booking CTA above fold to lift 1.2% → ~3.7% conversion · (3) Add FAQ schema to 3 pages for AI answer eligibility. All briefs are ready in the Execution tab.
⬡ Data Intake Agent · Stage 1 of 5
Last run: May 31, 2026 · 9:14 AM · All sources nominal
● 5 / 5 Connected
Google Analytics 4
Behavioral · Session Data
✓ Imported✓ Cleaned✓ Standardized✓ Ready
sessionsusersconversionsengagement_timebounce_rate
Google Search Console
Search Visibility Data
✓ Imported✓ Cleaned✓ Standardized✓ Ready
queriesimpressionsctrpositionpages
Meta Social Analytics
Social · Engagement Data
✓ Imported✓ Cleaned✓ Standardized✓ Ready
reachengagementimpressionsroasctr
Website Performance Data
Technical · Core Web Vitals
✓ Imported✓ Cleaned✓ Standardized✓ Ready
lcpclsfidpage_speed
Keyword Intelligence Data
SEO Findings · Competitive Rankings
✓ Imported✓ Cleaned✓ Standardized✓ Ready
keyword_rankingssearch_volumegap_analysiscompetitor_dataserp_features
⬡ Insight Agent · Stage 2 of 5
12 insights detected across 5 data dimensions · May 2026
High-Impression / Low-CTR Queries
QueryImpr.CTRGap
botox savings near me12,4001.9%-4.3%
med spa chicago9,1002.1%-2.9%
dermal filler cost7,8001.4%-3.6%
lip filler near me5,6003.1%-1.4%
Non-Branded SEO Findings
Top Traffic Sources
Content Performance Signal
↑ High Performing
patient-reviews 4.4% CTR
med-spa-services 3.1% CTR
before-after-gallery 2.9% CTR
↓ Underperforming
botox-savings 1.9% ↓
dermal-fillers 1.4% ↓
about-us 0.8% ↓
Weekly Session Trend — 8-Week View
Apr 7Apr 14Apr 21Apr 28May 5May 12May 19May 26
⬡ Strategy Agent · Stage 3 of 5
5 opportunities scored by impact × confidence ÷ effort
SEO Opportunity#1 Priority
CTR Recovery — High-Impression Keyword Cluster
12,400 monthly impressions for "botox savings near me" yielding only 1.9% CTR vs 6.2% expected. Title tag mismatch and missing price signals are suppressing clicks. Conservative recovery: +520 clicks/month.
Impact
9/10
Effort
2/10
Confidence
High
Conversion Opportunity#2 Priority
Booking Flow Friction Reduction
4m 48s average session, 1.2% conversion — 3.1× below industry median. Users are engaged but not converting. CTA placement, form complexity, and unclear next steps are primary friction points.
Impact
10/10
Effort
5/10
Confidence
High
Local Visibility#3 Priority
Local SEO Content Gap — Chicago Med Spa
Competitors rank for 14 high-intent local queries where this practice has zero presence. "Med spa Chicago," "aesthetic clinic near me," and "botox specials Chicago" are uncontested with strong commercial intent.
Impact
7/10
Effort
5/10
Confidence
Med
Content Gap#4 Priority
AI Answer Visibility — FAQ & Schema Gap
Zero appearances in AI-generated overviews for any target query. Competitors with structured FAQ schema are cited in Perplexity and Google AI Overviews. E-E-A-T signals are missing across top service pages.
Impact
7/10
Effort
3/10
Confidence
High
#5 Priority
IG Reels Scale — 3.8× Performance Gap
Reels outperform all other content types by 3.8× but comprise only 18% of the posting schedule. Shifting to 60% Reels focused on patient conversion could 2× overall social reach in 60 days.
Impact
6/10
Effort
4/10
Confidence
High
⬡ Execution Agent · Stage 4 of 5
5 implementation briefs generated · All actions ready to assign
Critical
Rewrite title tags on top-3 underperforming pages
Pages ranking 3–5 with CTR 2–4× below expected. Title tags lack price signals, urgency, and local modifiers users are actively searching for.
Rewrite /botox-savings to "Botox from $X · Chicago Med Spa". Mirror pattern across /dermal-fillers and /med-spa-services. A/B test urgency modifier in meta description.
Critical
Move CTA above fold, reduce booking form to 3 fields
Users spending 4m 48s on-site but converting at 1.2% — 3.1× below median. Primary booking CTA is below the fold on all service pages.
Move booking CTA above fold on all service pages. Simplify form: Name, Phone, Service. Add review count (social proof) adjacent to every CTA button.
High Priority
Add FAQ schema to top 3 pages for AI answer eligibility
Zero AI Overview appearances for any target query. Structured FAQ markup is the fastest path to AI citation eligibility and featured snippet capture.
Add 5-question FAQ blocks on /botox-savings, /dermal-fillers, /med-spa-services. Implement JSON-LD FAQ schema. Target: "How much does X cost in Chicago?"
Medium Priority
Publish local landing page targeting Chicago med spa cluster
14 high-intent local queries uncontested. A dedicated local page captures this traffic without cannibalizing existing rankings.
Draft /chicago-med-spa targeting "med spa Chicago" (9,100 impressions). Include LocalBusiness schema, map embed, 3 neighborhood testimonials.
Opportunity
Shift social calendar to 60% Reels, focus on patient conversion
IG Reels drive 7.1% ER vs 0.4% for static images. Current schedule is only 18% Reels — the highest-ROI social action available.
3 Reels / 1 Carousel / 1 Story Poll per week. Topic: before/after results, patient stories, price transparency. Retire low-ER educational static posts.
⬡ Evaluation Agent · Stage 5 of 5
Executive report generated · Pipeline complete
InsightRx
Executive Marketing Intelligence Report
Opportunity Score
5.37
Key Findings
Critical CTR Gap: 12,400 monthly impressions for top keyword cluster yielding 1.9% CTR vs 6.2% benchmark — ~520 clicks/month unrealized from title tag mismatches alone.
Conversion Underperformance: Strong engagement (4m 48s, 3.2 pages/visit) not translating to bookings — 1.2% conversion rate is 3.1× below industry median.
Local Search Gap: Zero ranking presence across 14 high-intent local queries where competitors are established. Uncontested traffic available now.
AI Visibility Zero: No appearances in Google AI Overviews or Perplexity for any target query. Competitors with FAQ schema are being cited instead.
Social Content Arbitrage: IG Reels outperforming all formats by 3.8× but only 18% of schedule. High ROI reallocation available at low incremental cost.
Strategic Recommendations
01Rewrite title tags on top 3 pages with price signals and local modifiers → +520 clicks/month
02Move booking CTA above fold and simplify form to 3 fields → target 3.7% conversion
03Add FAQ schema to 3 service pages to qualify for AI answer citations
04Publish Chicago local landing page with LocalBusiness schema for 14 uncontested queries
05Shift social calendar to 60% Reels focused on patient conversion narratives
Expected Impact
+520
organic clicks/month from CTR recovery
2–3×
conversion rate improvement potential
14
new local queries within reach
2×
social reach within 60 days
9:41●●●●
Opp. Score
5.37
↑ +0.4
CTR Gap
4.3%
Below benchmark
Traffic
+12%
vs prior avg
Agents
5/5
● Done
Session Trend
Pipeline Status
Data Intake ✓ → Insight ✓ → Strategy ✓ → Execution ✓ → Evaluation ✓ · All 5 agents complete
⬡ Insight Agent
47,210 data points analyzed · 12 insights detected
Top Traffic Source
google/organic 68.3% · direct 14.1% · instagram 9.7%
High CTR Pages
patient-reviews 4.4% · med-spa-services 3.1%
Low CTR Pages
botox-savings 1.9%↓ · dermal-fillers 1.4%↓
Keyword Impression Volume
⬡ Execution Agent · 5 Briefs Ready
All actions implementation-ready
🔴 Critical
Rewrite title tags — 3 pages
Add price signals + local modifiers.
Est. +520 clicks/month · 1–2 days
🔴 Critical
Move CTA above fold
1.2% conversion → target 3.7%.
Simplify form to 3 fields · 1 week
🟡 High
Add FAQ schema to 3 pages
Qualify for AI Overview citations.
2–3 days · High confidence
🟢 Opportunity
Shift to 60% IG Reels
3.8× ER vs static — scale this.
Ongoing · Low effort
⬡ Evaluation Agent
Opportunity Score: 5.37 · Report complete
Clicks/mo
+520
CTR recovery
Conv. Lift
2–3×
potential
Local Gaps
14
uncontested
Social Reach
2×
60 days
Priority Actions
1. Rewrite title tags · 2. Move CTA above fold · 3. Add FAQ schema · 4. Local landing page · 5. Scale IG Reels
Reflection
From Campaign Thinking
to Systems Thinking
Building InsightRx in a production environment changed how I think about marketing, AI, and the intersection of both.
How my thinking shifted
Campaign Thinking
→
Systems Thinking
Shifted from “what should I do?” to “how do I architect a system that answers that automatically?” Every strategy now looks like an optimization problem waiting to be built.
Prompt Engineering
→
AI System Design
Prompts alone are not enough. Production AI requires grounded data, structured outputs, agent orchestration, and evaluation frameworks. Prompts are one component of a larger system.
SEO
→
SEO + AEO + GEO
After auditing real brands against ChatGPT, Perplexity, and AI Overviews, it became clear: brands are already winning and losing in AI-mediated discovery — most don’t know it yet.
Strategy
→
Strategy + Engineering
The best strategy is only as useful as the system that can execute it. The best pipeline is worthless without real strategic questions. InsightRx forced me to hold both simultaneously.
What production AI actually requires
Prompts are necessary — but not sufficient.
InsightRx required all seven layers to produce reliable, grounded output.
Grounded Data
Structured Outputs
Agent Orchestration
Evaluation Frameworks
Data Normalization
Reliable Workflows
Prompt Engineering ↑ one of seven
Key takeaways
Grounding is the hardest part of AI
Getting agents to produce useful output is easy. Getting them grounded in live, clean, normalized data — handling authentication, rate limits, and schema mismatches — is where most AI projects fail.
data_groundingAEO and GEO are already real
Brands are already winning and losing in AI-mediated discovery — they just don’t know it yet. The window to build early advantage is narrow, and InsightRx is designed to identify it.
aeo_geo_visibilityStrategy and engineering are inseparable
InsightRx forced me to hold both simultaneously — strategic marketing questions grounding an AI pipeline, an AI pipeline making strategy executable. That intersection is where I want to work.
systems_thinkingFuture Direction
InsightRx
Development Roadmap
InsightRx is a working prototype with a clear path to production. Each phase extends the agent architecture, data coverage, and intelligence depth.
PHASE 01
Foundation
✓Six-agent pipeline architecture
✓GA4, GSC, search intelligence, Meta APIs
✓LangGraph A2A orchestration
✓Streamlit dashboard v1
✓Recommendation output layer
PHASE 02
Intelligence Depth
◐AEO citation monitoring (Perplexity)
◐GEO brand presence tracking
◐Competitor agent integration
◐Enhanced impact scoring model
◐Historical trend memory
PHASE 03
SaaS Architecture
○Multi-brand workspace support
○Scheduled pipeline runs
○Email/Slack recommendation delivery
○User authentication & billing
○API access for agency clients
PHASE 04
M2M Intelligence
○Agent-to-agent brand evaluation
○AI ecosystem brand scoring
○Predictive opportunity forecasting
○M2M readiness audit product
○Commercial launch
Get In Touch
Interested in InsightRx?
Let's Connect.
Whether you're a recruiter, a professor, a potential collaborator, or a brand that wants early access — I'd love to hear from you.