TrendzPulse
TrendzPulse
Role: Salesforce Admin & Developer
Technologies: Apex (Batchable, Schedulable, REST) | Flow Builder | Agentforce & Prompt Builder | Experience Cloud | Custom Objects | Guest User Security | YouTube Data API | Gemini | SendGrid | Telegram Bot API
I built a complete trend-intelligence product for YouTube creators, run entirely on Salesforce. Every Sunday a scheduled Apex engine pulls what is rising in a creator's niche, runs it through AI, and sends an interpreted digest by email and Telegram. Creators can also message a Telegram bot for content ideas on demand, powered by Agentforce.
YouTube creators struggle to keep up with what is trending in their niche. Existing tools hand them raw analytics and leave the interpretation to them, which means hours of research to answer one question: what should I make next? The need was for trend intelligence that arrives already interpreted, on a schedule, without the creator having to dig for it.
A complete trend-intelligence product built entirely on a free Salesforce Developer org, with no external server and no paid tooling.
The Engine (Automated)
A scheduled, batchable Apex class that runs every Sunday, one niche per transaction
Pulls trending videos from the YouTube Data API, has Gemini analyse them, and builds a branded HTML digest
Delivers by email through SendGrid and by Telegram, logging every send
The Creator Experience
A public Experience Cloud site for signup and niche selection
A branded welcome email on signup
A weekly digest covering what is rising, why it works, and one angle to film that week
The Conversational Layer
A Telegram bot creators connect by sending the email they signed up with
An Agentforce assistant that answers their content questions from their own niche data
A public Apex REST webhook that receives and routes every bot message
Delivers interpreted trend intelligence weekly, not raw data
Runs entirely on its own schedule with no manual intervention
Lets creators ask an AI assistant for ideas on demand, through a channel they already use
Captures every signup, delivery, and interaction in Salesforce as the source of truth
Objects: 4 custom objects (Creator, Digest Log, and two Telegram inbox objects)
Automation: 7 Flows (3 record-triggered, two of them calling Apex actions; 3 screen flows; 1 autolaunched feeding an Agentforce agent)
Apex: a scheduled/batchable digest engine, an invocable email class, and a public REST webhook
Agentforce: 2 Agentforce agents, a creator-facing assistant on a prompt template and an internal ops agent
Integrations: YouTube Data API, Gemini, SendGrid, Telegram Bot API
Platform: Experience Cloud public site with secure guest-user access
Salesforce-first, no external server: The entire engine runs in Apex inside the org instead of a paid automation tool, removing every external dependency.
Guest-user inbox pattern: A public bot runs as a guest user, which is restricted from reading existing records. The fix: the guest only creates a record, and a flow running in the background does the lookup and the real work. Secure, and reused across the whole bot.
AI picks, Apex builds: Gemini chooses videos by number while Apex constructs every URL and value, which keeps the model reliable and stops it inventing data.
One niche per batch transaction: Keeps each run within governor limits and isolates any failure to a single niche.
Suggestions, not claims: The assistant is scoped to content ideas rather than live-trend facts, so it stays useful without fabricating data.