Case study · Sales prospecting

Market intelligence that turns itself into a pipeline of qualified prospects, every morning.

Deployed for an influence agent and their artist roster, this system turns an entire sector of trade press into a stream of prioritized opportunities. Every morning it reads everything published in the field, discards the noise, ranks what remains and spots the brands that keep coming back, without a single line typed by hand. The user opens an already-sorted table and an already-written report.

25 feeds

Sources monitored continuously

Daily

Execution cadence

3 tiers

Priority levels

0 clicks

Manual intervention

01 The pipeline, from source to prospect

Five steps orchestrated in a chain, triggered every morning. Each produces a clean deliverable consumed by the next. No human handover, no copy-paste.

Collection

25 industry feeds gathered every morning, with full-text retrieval for every article.

AI analysis

Every article is read by a language model that keeps what is actionable and discards what is off-topic.

Prioritization

Each retained signal is ranked Tier 1, 2 or 3 against four business relevance criteria.

Brand register

Every brand accumulates run after run: repetition surfaces the hot prospect that no single article reveals.

Delivery

A table fed line by line and a daily report sent by email, even on days with no signal.

Collection

Twenty-five industry sources (trade press and monitoring queries) are polled on a scheduled trigger. Nothing is typed by hand: articles arrive in a normalized, dated format, and the full text is retrieved page by page, not just the headline.

Analysis

Articles are processed in batches by a language model, which lets it compare sources against each other and deduplicate a single campaign quoted by several outlets. Every retained signal comes out with a short, verifiable justification.

Prioritization

The safeguard against noise is not a tight filter, it is the tier. The system keeps a wide net and ranks: the user sees their Tier 1 items first, without ever risking the loss of a weak but real signal.

A dedicated tab keeps one line per brand, accumulated over time (occurrences, first and last detection, best tier). A brand seen five times in three weeks in a target sector is a prospect, even if no single article shouted it.

Delivery

A report goes out every day, even with zero signals, to remove any ambiguity between "nothing to report" and "breakdown". A separate alert fires if a feed dies or the collection runs empty.

02 What keeps the monitoring reliable over time

The real risk of automated monitoring is not the visible crash, it is silent degradation: sources dying one by one while the system reports "all good". The chain is designed to make that impossible.

Failure-proofing

Three safeguards raise an explicit error if no source responds, if content can no longer be parsed, or if nothing enters the current day's window, the scenario of the run that "succeeds" by writing nothing. Every failure sends a readable diagnosis by email.

Zero duplicates

Deduplication is twofold: within a run (two sources, one article) and across runs (already-known entries are never rewritten). A failing source is logged and skipped, it never breaks the collection.

Controlled cost

One pass per day, in optimized batches: a few euros of processing per month to cover an entire press sector, where equivalent human monitoring would take several hours every day.

03 The technical stack involved

No proprietary building blocks, no locked-in dependencies. Everything stays maintainable in-house, brick by brick.

n8n orchestration Claude Sonnet 5 25 RSS & monitoring feeds Tier-based scoring Versioned brand register Daily report Dedicated failure alert Scheduled trigger

The expertise on display

Turning an information stream into decisions.

The hard part was not reading articles, it was making sure a very busy professional opens the tool every single day and finds immediate value in it. That means sorting without ever losing anything, making visible what only shows at the scale of time, and guaranteeing the system flags itself the day it weakens. This is the kind of orchestration, reliable and built for real-world use, that we know how to build and maintain for business processes of any kind.

Book a discovery call

A use case close to yours?

A free 20-minute diagnosis to find out whether the same mechanism applies to your business, no jargon, no commitment.

Book a slot