I Built an Analytics Dashboard with AI — No Subscriptions, No Analyst, No Code
I didn't just ask AI to build me a dashboard. My AI agent connected to Google Search Console, Supabase, and Meta Ads on its own — handled OAuth, authentication, and data queries — then built a unified dashboard, deployed it to Vercel, and now uses that data to make its own decisions. The dashboard isn't the product. The feedback loop is.
I replaced all of it with an AI agent that built its own analytics dashboard, deployed it for free, and now autonomously monitors the data to inform its own decisions about my content strategy, ad spend, and growth priorities. No subscriptions. No analyst. No code.
The Analytics Problem Nobody Talks About
Let me describe a morning that probably sounds familiar.
You open Google Analytics. There are 47 different reports. You click around for ten minutes, look at a traffic graph that goes up and to the right, feel vaguely good about it, and close the tab. Then you open your email platform. Different numbers. Different time ranges. Different definitions of what a "conversion" even means. Then maybe you check your ad platform. More numbers. More graphs. More confusion.
Thirty minutes later, you still can't answer the only question that matters: Is what I'm doing actually working?
This was my life for longer than I'd like to admit. I had Google Analytics, Google Search Console, a Supabase database for my CRM, Meta ads running, an email platform — each one with its own dashboard, its own login, its own version of the truth. And I'm a growth analyst by trade. I'm supposed to be good at this.
The problem isn't that we lack data. The problem is that we're drowning in it. Modern analytics tools are built for enterprise teams with dedicated data analysts. They show you everything because they don't know what matters to your specific business. So they give you 100 metrics and let you figure it out.
Most business owners respond to this in one of two ways. Either they ignore analytics entirely ("I'll just focus on the work and hope it grows") or they become data hoarders — checking seven dashboards daily without ever making a decision based on what they see.
Both approaches leave money on the table.
What I actually needed was dead simple: one screen, five numbers, updated automatically, that I could check in 60 seconds every morning and know exactly where my business stood. That's it. Not 100 metrics. Not seven dashboards. Five numbers.
But here's what I didn't expect: the dashboard wasn't the real breakthrough. The real breakthrough was that the AI agent that built the dashboard also uses it. It doesn't just display the data for me — it reads that data to make its own decisions about what to do next. More on that in a minute.
The 5 Metrics My Agent Monitors Autonomously
Before I talk about the dashboard itself, let me explain which numbers matter — and how the agent uses each one.
After years of doing growth for businesses — and now running my own — I've narrowed it down to five metrics. These are the five numbers my AI agent monitors autonomously. When something changes, it alerts me. When something needs attention, it acts.
1. Daily Website Visitors
What it tells you: Are people finding you?
This is your top-of-funnel health check. If this number is growing, your SEO, content, and distribution are working. If it's flat or declining, nothing downstream matters — you can't convert visitors you don't have.
I track this from Google Search Console rather than Google Analytics because GSC shows me search traffic specifically — the people actively looking for what I offer. That's the traffic that converts.
How the agent uses it: When SEO traffic dropped on a Sunday, the agent pulled the GSC data, analyzed it, and explained the weekend pattern — business audiences don't search on weekends, so a Sunday dip is normal, not a crisis. It told me to compare Tuesday-to-Tuesday instead of stressing over daily swings. That's the kind of context that saves you from panic-driven bad decisions.
2. Email Signups
What it tells you: Are visitors interested enough to raise their hand?
Traffic without signups is just window shopping. This number tells you whether your offer, your lead magnet, and your messaging are compelling enough that people want to hear more from you. It's the bridge between "stranger who found your site" and "potential customer."
How the agent uses it: It queries the Supabase CRM for signup counts and calculates conversion rates against traffic. When the rate dropped below 1%, the agent flagged it and suggested the lead magnet was misaligned with visitor intent — before I even noticed the problem.
3. Ad Spend vs. Leads
What it tells you: Is your paid acquisition profitable?
If you're running ads, this is the number that keeps you from burning cash. Not impressions. Not clicks. Not CTR. How much did you spend, and how many actual leads did it generate? That gives you your cost per lead, which you can compare against how much a customer is worth to you.
How the agent uses it: When ad CTR looked good on our Meta campaigns, the agent didn't celebrate — it flagged that traffic campaign CTR doesn't predict conversion performance. Good click-through on a traffic objective tells you the ad is interesting. It tells you nothing about whether those clicks become leads or customers. The agent caught the distinction I might have missed.
4. Top-Performing Content
What it tells you: What's actually driving traffic and interest?
Most business owners create content and hope for the best. This metric shows you which pages, posts, or articles are doing the heavy lifting — so you can double down on what works instead of guessing. I track my top 10 pages by clicks from search. The pattern always surprises me.
How the agent uses it: It checks GSC data to inform content strategy — identifying which topics are gaining traction and where to focus next. When it spotted posts ranking on page 2 for high-intent keywords, it added them to the content optimization queue without me having to ask.
5. Revenue
What it tells you: Is all of this activity turning into money?
This is the number that keeps you honest. Traffic can grow, signups can climb, ad costs can look great — but if revenue isn't following, something in your funnel is broken. I track total revenue, broken down by source (organic vs. paid vs. email), so I know which channels are actually making money.
How the agent uses it: Revenue per channel helps the agent recommend where to invest. When organic was outperforming paid on a revenue-per-visitor basis, the agent recommended shifting more budget toward content production and less toward ads. Data-driven allocation, not gut feel.
That's it. Five numbers. But here's the difference between a regular dashboard and what I have: I don't have to remember to check. The agent monitors these autonomously. When something changes — traffic drops, signups spike, ad costs creep up — it alerts me. And when I ask "what should we focus on this week?", it already knows the answer because it's been watching the numbers the whole time.
What the Agent Actually Did (Not "Built a Webpage")
When I say "I built an analytics dashboard with AI," most people picture someone typing a prompt and an AI spitting out some HTML. That's not what happened. What happened was a multi-step autonomous operation where the agent acted more like a systems integrator than a code generator.
Here's the actual sequence of what my AI agent did, step by step:
Step 1: Connected to Google Search Console API
The agent set up OAuth credentials in Google Cloud Console, handled the authentication flow, created the service account, and configured the permissions so it could query my search data programmatically. This isn't filling in a form — it's navigating a multi-step API credential setup that trips up most developers.
Step 2: Connected to Supabase
The agent connected to my Supabase database — the same one that powers my CRM with 8 tables, row-level security, and edge functions. It queried lead data, email signup records, and customer information. It already knew the schema because it built the database in a previous session.
Step 3: Connected to Meta Ads API
The agent set up the Facebook app credentials, obtained a long-lived access token, connected to my ad account, and pulled campaign performance data — spend, impressions, clicks, CTR, and leads. Three separate API integrations, each with their own authentication dance.
Step 4: Built a Unified Dashboard
With all three data sources connected, the agent built a single-page application that pulls from all of them simultaneously. Not three iframes. Not a mashup of screenshots. One unified interface showing five clean sections: traffic, signups, ads, content, and revenue — each pulling live data from the appropriate source.
Step 5: Deployed to Vercel
The agent pushed the code to git, which triggered an automatic deployment to Vercel. Live on a URL within minutes. Set up password protection so only I can access it.
Step 6: Started Using the Data
This is the part that makes it genuinely agentic — not just "AI built a thing." The agent now uses this dashboard data in its own autonomous operations. When I ask it about content strategy, it checks the GSC data first. When I ask about ad performance, it pulls the latest Meta numbers. When I ask "what should we do this week?", it has real data informing its recommendations, not generic advice.
The whole thing loads in under two seconds, is password-protected so only I can see it, and lives at a URL I can bookmark on my phone. No app to install. No subscription to manage. No analytics vendor trying to upsell me to an enterprise plan.
How It Pulls Real Data (Explained Simply)
One of the most common questions I get is: "How does the dashboard actually get the data?" It sounds technical, but the concept is straightforward.
Think of it like plumbing. Your data lives in different places — Google has your search data, your database has your signups, Meta has your ad data. The dashboard is like a faucet that connects to all those pipes and shows you the water flowing through them.
In slightly more specific terms:
Google Search Console has an API — basically a door that lets other software ask it questions. My dashboard asks it: "How many clicks did I get today? What are my top pages?" Google answers, and the dashboard displays it. The agent set up the credential file, created the service account, and configured the permissions — the kind of fiddly multi-step process that would take a human developer 30-60 minutes of reading documentation.
Supabase is where I store my business data — email signups, customer records, revenue. It's like a spreadsheet in the cloud, but more powerful. The dashboard connects to it directly and pulls the numbers I care about. The agent already knows my database schema because it built the 8-table CRM in an earlier session — so it knew exactly which tables to query and how to join the data.
Meta (Facebook/Instagram) ads data comes through Meta's Marketing API. Same concept — the dashboard asks Meta "how much did I spend today and how many leads did I get?" and displays the answer. The agent handled the Facebook app creation, token generation, and ad account connection autonomously.
The AI agent set up all of these connections. I didn't write code. I didn't read API documentation. I didn't debug authentication errors. The agent handled the entire integration pipeline — and when something didn't work (the Google credential setup was fiddly), it diagnosed the error and fixed it on its own.
The entire dashboard is hosted on Vercel — a platform that hosts websites for free. My dashboard costs exactly $0/month to run. The data sources (Google Search Console, Supabase's free tier, Meta's API) are all free too.
Let me be blunt about what this replaces: a comparable setup using off-the-shelf analytics tools — something like Mixpanel or Amplitude for product analytics, plus Databox or Klipfolio for the dashboard layer, plus whatever connectors you need to pull data from multiple sources — would run $500-1,000/month. And you'd still be looking at someone else's template instead of exactly the five metrics you care about.
The Feedback Loop: When the Agent Uses Its Own Dashboard
Here's where this stops being a "cool AI project" and starts being a fundamentally different way to run a business.
My AI agent doesn't just build things and walk away. It operates continuously — checking data, making recommendations, catching problems, and informing its own decisions with real numbers. The dashboard it built is now part of its own decision-making infrastructure.
Let me give you three real examples from this week:
🔍 Example 1: The Sunday Traffic "Drop"
SEO traffic dropped on Sunday. My instinct would have been to worry. Instead, the agent pulled the GSC data, compared it to previous Sundays, and explained the pattern: business-focused content naturally sees lower weekend search traffic. It told me to track week-over-week trends on the same weekday, not daily swings. A simple insight — but one I might have missed in a moment of data-driven panic, and one the agent could make because it had the data.
📊 Example 2: Good CTR, Unknown CPL
Our Meta ads were showing strong click-through rates. Looks great, right? The agent flagged something most marketers overlook: we were running traffic campaigns, not conversion campaigns. A good CTR on a traffic objective means the ad creative is interesting. It says nothing about whether those clicks convert to leads — and the cost per lead was unknown because the campaign wasn't optimized for conversions. The agent recommended restructuring the campaign before we drew any conclusions about ROI.
📈 Example 3: Content Strategy from Real Data
When I asked the agent what content to write next, it didn't brainstorm from thin air. It pulled the latest GSC data, identified which posts were gaining impressions but underperforming on clicks (meaning the topics had demand but our titles or meta descriptions weren't compelling enough), and recommended specific updates. It also identified keyword gaps — topics competitors were ranking for that we hadn't covered. Strategy built on data, not guesswork.
This is the feedback loop that changes everything: the agent builds the analytics → monitors the analytics → uses the analytics to make better decisions → those decisions generate new data → repeat.
Most people think of AI as a tool that does what you tell it. This is AI as a system that observes, learns, and adapts. The dashboard isn't a destination — it's one component in a larger intelligence loop that gets smarter the more data flows through it.
The Honest Downsides
I'm not going to pretend this is perfect. There are real trade-offs.
Despite all this, I'd make the same choice again in a heartbeat. The 3 hours of setup time has saved me 30+ minutes every single day. And the real payoff isn't the dashboard — it's having an agent that understands my business data and uses it to make better decisions on my behalf.
How to Build Your Own AI Analytics Dashboard
If you want to build something similar, here's the approach I'd recommend. You don't need any technical background. You need to know what you want to track and where your data currently lives.
Step 1: Define Your Five Metrics
Before you touch any tools, write down the five numbers you'd check every morning if you could only pick five. Use my list as a starting point, but customize it for your business:
- Service businesses: Website visitors, inquiry form submissions, proposal close rate, active clients, monthly revenue
- E-commerce: Store visitors, add-to-carts, orders, average order value, revenue
- SaaS: Signups, activation rate, active users, churn, MRR
- Coaches/consultants: Website visitors, discovery call bookings, close rate, active clients, revenue
The right five metrics follow the journey: Are people finding me → Are they interested → Are they buying → Am I making money?
Step 2: Identify Where Your Data Lives
For each metric, write down where the data currently sits:
- Website traffic → Google Search Console or Google Analytics
- Email signups → Your email platform (Mailchimp, ConvertKit, or your database)
- Ad performance → Meta, Google Ads, LinkedIn Ads
- Sales/revenue → Stripe, your CRM, your database
- Content performance → Google Search Console
Step 3: Give the AI a Clear Brief
This is the actual prompt. Customize the bracketed parts for your business:
The Dashboard Prompt
Build me a single-page analytics dashboard with these sections:
1. TRAFFIC: Daily visitors from [Google Search Console / Google Analytics]
for the last 30 days, with a chart and week-over-week comparison.
2. SIGNUPS: New [email signups / leads / inquiries] from [your database
or email platform], today + this week + this month, with conversion
rate (signups / visitors).
3. ADS: Total spend vs. leads from [Meta / Google Ads] this week and
this month, with cost per lead and a green/red indicator vs. my
target of $[X] per lead.
4. CONTENT: Top 10 pages by search clicks from Google Search Console,
with impressions and average position.
5. REVENUE: Total revenue from [Stripe / your database] broken down
by source, this week and this month.
Requirements:
- Password-protected (simple password gate is fine)
- Hosted on Vercel (free tier)
- Mobile-friendly
- Loads real data from APIs, not dummy data
- Clean, minimal design — I want to scan this in 60 seconds
Connect to each data source autonomously. Set up OAuth credentials,
handle authentication, and pull live data. Then deploy it and use
the data in your ongoing analysis of my business.Step 4: Let the Agent Handle the Integrations
This is where a real AI agent differs from a code generator. The agent will:
- Google Search Console: Create the service account, download credentials, configure permissions, test the connection
- Your database: Connect using your API URL and key, query the right tables, handle the data joins
- Meta Ads: Set up the Facebook app, generate a long-lived access token, connect to your ad account
- Stripe: Connect via API key, pull revenue data by source
Each integration takes 10-30 minutes. The agent handles the technical implementation and troubleshoots errors along the way. You provide credentials and approve connections — the agent does the rest.
Step 5: Deploy and Close the Loop
Once everything's working, the agent deploys it to Vercel with a single command. You get a URL. Bookmark it. But more importantly — the agent now has access to all this data for its ongoing operations. When you ask it about strategy, it pulls real numbers. When something changes, it can alert you. The dashboard is just the visible surface of a much deeper data integration.
Total time from "I want a dashboard" to "my agent is monitoring my metrics autonomously": one afternoon. Most of that time is the API credential setup, not building the dashboard itself.
Frequently Asked Questions
Q: Can AI really build a working analytics dashboard?
Yes — but "build" understates what actually happens. A modern AI agent doesn't just generate HTML. It autonomously connects to APIs like Google Search Console, Supabase, and Meta Ads — handling OAuth credentials, authentication flows, and data queries — then builds the unified interface, deploys it, and continues using the data in its own decision-making. It's closer to hiring a systems integrator than asking for a webpage.
Q: What does an AI analytics dashboard cost to run?
The dashboard itself is hosted for free on Vercel. The data sources I connect to (Google Search Console, Supabase free tier, Meta's API) are all free. The only cost is the AI tool you use to build and run it. Total ongoing hosting cost: $0/month. Compare that to $500-1,000/month for a comparable analytics stack built from off-the-shelf tools.
Q: What metrics should a business owner track daily?
Focus on five numbers: daily website visitors (are people finding you?), email signups (are they interested enough to opt in?), ad spend vs. leads (is your paid acquisition profitable?), top-performing content (what's actually driving traffic?), and revenue (is all this activity turning into money?). Everything else is noise until these five are healthy. And ideally, have an AI agent monitoring them so you get alerts when something changes, not just charts to stare at.
Q: Do I need coding skills to build an AI analytics dashboard?
No. The agent handles the entire technical implementation — connecting APIs, writing code, deploying the application, setting up authentication. You need to know what you want to track and where your data lives. The agent handles everything else, including troubleshooting its own errors along the way.
Q: What if I don't use the same tools (Supabase, GSC, Meta)?
The approach works with whatever tools you use. If your email list is in Mailchimp instead of Supabase, the agent connects to Mailchimp's API. If you run Google Ads instead of Meta, same principle. The dashboard is custom-built for your stack — that's the whole point. Just tell the agent what you use and where your data lives.
Q: How long does the dashboard stay accurate?
It pulls live data every time you load it, so it's always current. The only thing that would break it is if one of your data sources changes their API (rare) or your credentials expire (Google tokens need refreshing occasionally). If something breaks, the agent diagnoses the error and fixes it — usually in under five minutes. And because the agent uses this data in its own operations, it notices breakages quickly.
Free: The AI Growth Breakdown
See how one business went from 0 to 100+ daily visitors in 14 days using AI agents. The exact tools and results.
Get the Free Breakdown →