Case Study

Monday mornings,
automated.

Every Monday at 9am, a structured sector briefing lands in the client's workspace. Industry news, competitor moves, and actionable ideas with embedded domain knowledge.

Status Live · Client Delivered
Client Paris Architecture Firm
Category Scheduled Research
Built 2026
At a glance

What is Automated Sector Research Briefing?

The Automated Sector Research Briefing is a scheduled Claude API system built for a Paris-based architecture firm. Every Monday at 9am it pulls industry news, competitor moves, and actionable ideas from curated sources, reasons about them using an embedded architecture knowledge document, and delivers a structured briefing that reads like an expert wrote it. Replaced two hours of manual research every week.

The Problem

What was broken.

The firm\'s director had a standing Monday ritual: two hours reading architecture news, competitor blogs, and design publications before the weekly team meeting. The point was not the reading itself. It was arriving at the meeting with three or four sharp angles on what was happening in the industry that week, so the team could discuss them.

Generic AI summarisation tools produced summaries that sounded like AI summaries. Generic news aggregators produced lists of headlines with no synthesis. Neither replaced the actual job, which was to read with a trained eye and pull out the ideas worth talking about. The output had to reason like an architect, not a generalist.

The two hours was the constraint. Monday morning is the worst time to lose two hours. The director needed the briefing to exist without them having to produce it.

The Approach

What was built.

I built a scheduled Claude API pipeline that runs every Monday at 8am Paris time. The system has three layers: a source-fetching layer that pulls from a curated list of architecture publications and competitor sites, a reasoning layer that uses Claude with an embedded architecture knowledge document to analyse what is actually newsworthy, and a delivery layer that writes the final briefing to the firm\'s internal workspace.

The key to making the output read like an architect wrote it was the knowledge document. I spent a full discovery session building a structured brief of the firm\'s architectural vocabulary: the movements they follow, the concerns that matter (embodied carbon, climate adaptation, urbanism), the competitors they care about, the publications they trust. Claude reads this document before it reads the news every Monday.

The result is a briefing that knows what "good" looks like for this specific firm, not a generic news roundup.

How It Works

Architecture in plain English.

01
Source fetching
Every Monday at 8am, the pipeline pulls the latest articles from a curated list of architecture sources, competitor websites, and industry blogs. Uses Firecrawl for reliable content extraction.
02
Knowledge embedding
Before analysing anything, Claude loads the firm's architecture knowledge document: vocabulary, movements, competitors, concerns. This is the layer that makes the output industry-specific.
03
Relevance reasoning
Claude reviews every source against the knowledge document and picks the 5-8 items that actually matter to this specific firm. Filters out noise, promotional content, and generic trend pieces.
04
Briefing synthesis
Claude writes the final briefing: headline items, the "so what" for each, direct quotes where useful, and a "watch list" section for things developing but not yet decisive.
05
Delivery
The finished briefing lands in the firm's internal workspace by 9am Monday. Ready for the director to skim before the 10am team meeting.
Try It

See it in action.

Sample sector briefing

This is a sample. The live system pulls from curated sources every Monday.

Stack

Built with.

Python Claude API Claude Sonnet 4.6 Firecrawl cron (scheduled) Knowledge embedding Markdown output
Outcomes

What changed.

2h → 0h manual research per week
8am delivery every Monday
100+ hours saved per year
0 weeks missed since launch

The briefing is not meant to replace the director's judgement. It is meant to give them the raw material in a form they can act on, so Monday morning becomes a 10-minute skim instead of a 2-hour deep dive. That is the right use of AI: removing the grunt work that precedes the real thinking, not replacing the thinking itself.

Free Consultation

Want a system
like this one?

Book a free 30-minute call. We map your situation, scope a similar build, and agree on a fixed quote before anything starts.

Book Free 30-min Call