ContentTool

Internal Tool

Agentic AI content engine that plans, writes, and schedules a company's full marketing cadence

PythonFastAPISvelteKitTypeScriptGeminiGitHub Actions
A
Ad Hoc Data
Content
Workflow
Inspiratie
Archief
Evaluatie
Bronnen
Kalender
Versie 0.1.0 · MVP
Workflow
Maak je volgende contentstuk
Social postsBlogsNieuwsbrieven
LinkedInIngepland
Kunstmatige intelligentie en de toekomst van werk
De snelle evolutie van AI stelt organisaties voor nieuwe uitdagingen…
#AI #Werk #Toekomst #Innovatie
78
AI-boom sluit miljarden arbeidsplaatsen buiten
Telegraaf · 2 dagen geleden
OpinieKies
3/wk
Social posts
1/wk
Blog article
1/mo
Newsletter
Daily
Scrape & score run

Overview

ContentTool is a full-stack agentic system I built at Ad Hoc Data to run the company's content marketing on autopilot. Every day it scrapes industry sources into an inspiration pool, scores each item with Gemini, enriches promising angles with CBS open-data statistics, and drafts posts onto a content calendar — three social posts a week, a weekly blog, and a monthly newsletter.

The Challenge

Consistent, relevant B2B content takes hours of recurring manual work: finding inspiration, validating it against real data, writing in the company's voice, and keeping a steady cadence. Generic AI writing tools produce generic output because they lack the company's data context. The system needed to generate content that is grounded in actual market statistics and runs without anyone remembering to trigger it.

The Solution

Daily scrape-and-score pipeline

A GitHub Actions cron triggers the daily run: httpx, BeautifulSoup and feedparser collect from industry sources (Playwright handles JS-rendered pages), then Gemini scores every item for relevance and freshness before it enters the inspiration pool.

Data-grounded drafting

Promising inspiration is matched against CBS open-data statistics, so generated angles and hooks reference real numbers instead of generic claims. Gemini drafts the posts; the content calendar enforces the publishing cadence.

FastAPI backend, SvelteKit dashboard

Python 3.12 with FastAPI and SQLAlchemy 2.0 (Alembic migrations) powers the pipeline; a typed SvelteKit + Tailwind dashboard gives the marketing team review-and-edit control before anything ships.

Newsletter delivery via Brevo

The monthly newsletter is assembled from the month's best-performing content and delivered through Brevo — the whole loop from inspiration to inbox is automated.

Key Features

Daily Scraping

httpx + BeautifulSoup + Playwright for JS-heavy sources

Gemini Scoring

Every inspiration item scored for relevance before use

Agentic Drafting

Angles, hooks, and full drafts generated from scored input

Content Calendar

Steady cadence: 3 socials, weekly blog, monthly newsletter

CI-Driven

GitHub Actions cron runs migrate → seed → scrape → score

Brevo Delivery

Automated newsletter assembly and sending

Results

Live and operational for Ad Hoc Data's marketing

Full weekly content cadence maintained without manual planning

Content grounded in CBS statistics instead of generic AI claims

Human review built in: the team approves drafts from the dashboard