HARCHAtelier
Méthode · 3 étapes

Monitor. Analyze. Deliver.
Three steps, every day.

No magic. No black box. Just a clear pipeline: we collect every mention of your brand across media and AI engines, classify it with NLP + LLM, and deliver insights to your WhatsApp, dashboard, and monthly PDF.

30+
Media sources
8
AI engines
3
Languages
< 5min
Alert latency
Pipeline overview
01
Monitor
Collecte 24/7
02
Analyze
NLP + LLM
03
Deliver
WhatsApp + PDF
Daily output~247 mentions / day
How it works

Three steps. No black box.

Most reputation tools sell you a magic score. We show you the pipeline. Every insight we deliver can be traced back to a source article, an AI response, and a sentiment score with its confidence interval.

01

Monitor

Collecte 24/7

30+ sources media, 8 moteurs IA, 3 langues. Crawl continu, RSS, APIs, scraping éthique.

30+
sources
24/7
crawl
02

Analyze

NLP + LLM

Détection d'entités, sentiment, sujets, langue. Classification multi-axes par HarchIQ.

3
langues
HarchIQ
IA trainable
03

Deliver

WhatsApp + PDF

Daily digest 7h, alertes crise <5min, dashboard live, rapport mensuel PDF.

< 5min
alerte
7h00
digest
Pipeline diagram

From raw article to WhatsApp alert.

The full data pipeline. Every box is a process, every arrow is a data flow. Latency targets are shown on each transition.

COLLECTEINGESTIONANALYSELIVRAISONRSSAPIwebhookMedia sources30+ sitesAI engines8 LLMsSocial signalsRSS + APIsIngestionCrawl · Parse · DedupNEREntity extractionSentimentHarchIQ classifyTopic modelBERTopicWhatsApp digest7h00 dailyCrisis alerts<5min latencyDashboardliveMonthly PDFboard-ready~2min~1min~2min
STEP 01~2min latency

Monitor — we never miss a mention.

We crawl 30+ Moroccan and African media sources every 60 seconds. RSS, APIs, and ethical scraping. We also query 8 AI engines every hour to see what they say about you on the prompts that matter.

12 of 30+ media sources
Le MatinFR
L'ÉconomisteFR
HespressAR/FR
TelQuelFR
Médias24FR
Aujourd'hui Le MarocFR
Le DeskFR
ChallengeMAFR
L'OpinionFR
AssabahAR
YabiladiFR
Bladi.netFR
30+
Media sources
8
AI engines
60s
Crawl interval
3
Languages
atelier.harchcorp.com / monitor / live● Live
2,847
Crawled today
47
New (1h)
312
Dedup'd
8
Queued
Recent intake · last 60 minutes
12:42HespressBank of Africa: nouveau partenariatAR
12:31L'ÉconomisteMarché monétaire: la BAM injecte 8 MdsFR
12:18Médias24Telecom: Inwi dépasse 15M d'abonnésFR
11:59TelQuelOCP Group annonce un investissementFR
11:47Le MatinRAM: nouveaux vols vers l'AsieFR
11:32Aujourd'huiCIH Bank lance offre premiumFR
Last crawl: 12:47:03● All sources operational
atelier.harchcorp.com / analyze / hes_3489
Source: hespress.com · 12 mars 2025FR
Frais bancaires: les clients dénoncent une hausse

Sur les réseaux sociaux, de nombreux clients se plaignent d'une augmentation des frais de tenue de compte. Bank of Africa n'a pas encore réagi officiellement...

NER — Entities done
Bank of AfricaBMCEMaroc
Language done
FR (0.94)AR (0.04)Darija (0.02)
Sentiment (HarchIQ) done
-0.61Negative
Topic (BERTopic) done
frais_bancairesservice_client
Crisis alert● triggered
Sentiment < -0.5 on tracked entity● Triggered
Final score
-0.61 / -1.0
Confidence92%
STEP 02~1min per article

Analyze — every word, every mention, classified.

Each article goes through three NLP stages: entity extraction (NER) to detect which company / person / product is mentioned, sentiment classification via HarchIQ in French, Arabic, and English, and topic modeling (BERTopic) to cluster themes over time.

2.1
NER — Named Entity RecognitionspaCy + custom
Detects companies, people, products, locations. Multi-language.
2.2
Sentiment classificationHarchIQ
HarchIQ assigns a score from -1 (negative) to +1 (positive) with confidence.
2.3
Topic modelingBERTopic
BERTopic clusters articles into themes: 'frais bancaires', 'résultats', 'service client'.
2.4
Language detectionfastText
FR / AR / EN / Darija — handles code-switching common in Moroccan media.
Output record (JSON)
{
  "article_id": "hes_3489",
  "source": "hespress.com",
  "entities": ["Bank of Africa", "BMCE"],
  "sentiment": -0.61,
  "sentiment_label": "negative",
  "confidence": 0.92,
  "topic": "frais_bancaires",
  "language": "fr",
  "alert": true
}
STEP 033 delivery channels

Deliver — insights where you already are.

No new app to install, no dashboard you have to remember to check. Insights land in WhatsApp, in your dashboard, and in your inbox as a monthly PDF.

7h00 daily

WhatsApp digest

Every morning at 7:00 — yesterday's mentions, sentiment score, top topics, and any crisis alerts.

Harch Atelier · 7:00
Bank of Africa · 13 mars
Score: 78/100 ↑ +4.2
Articles: 12 · AI: 3 citations
● 1 alerte (frais bancaires)
live 24/7

Live dashboard

Full historical view — 30 / 90 / 365 days. Drill into any article, any topic, any source. Export to CSV.

30-day trend+20 pts
monthly

Monthly PDF report

12-page board-ready report — reputation score, top risks, competitor benchmark, recommended actions.

Harch Atelier · Mars 2025
Reputation Report
Bank of Africa · 32 pages
68% pos22% neu10% neg
End-to-end timeline

From publication to your inbox — in 5 minutes.

Here is what happens when a negative article about your brand is published on Hespress at 12:42. Every stage is timestamped and logged. You can audit the full trail.

T+0min
Article publié sur hespress.com
T+2min
Crawler détecte le nouvel article
T+3min
NLP extrait entités + langue
T+4min
HarchIQ classe sentiment: négatif (-0.61)
T+5min
Alerte WhatsApp envoyée au comms team
T+7h00
Inclus dans le daily digest
T+30j
Inclus dans le rapport mensuel PDF
5min
Detection → alert
7h00
Daily digest
24h
Dashboard update
30j
Monthly PDF
Sample output

What a real analysis record looks like.

Four real mentions of Bank of Africa, classified by our pipeline. This is what you see in your dashboard every morning.

atelier.harchcorp.com / mentions / bank-of-africa
AllPositive (1)Neutral (1)Negative (1)
L'Économiste
12 mars 2025
Bank of Africa accélère son expansion en Afrique de l'Ouest
Le groupe bancaire marocain annonce l'ouverture de 12 agences au Sénégal et en Côte d'Ivoire...
positive
+0.82
Hespress
11 mars 2025
Frais bancaires: les clients dénoncent une hausse
Sur les réseaux sociaux, de nombreux clients se plaignent d'une augmentation des frais de tenue de compte...
negative
-0.61
Médias24
10 mars 2025
Bank of Africa publie ses résultats semestriels
Le bénéfice net s'établit à 1,2 milliard de dirhams, en hausse de 3% par rapport à 2024...
neutral
+0.08
TelQuel
09 mars 2025
L'application mobile de Bank of Africa élue meilleure du secteur
Lors de la cérémonie des Maroc Tech Awards, l'application BOA Mobile a remporté le prix...
positive
+0.71
Coverage

What we monitor, in numbers.

A complete picture requires complete coverage. Here is the full list of what our pipeline touches every day.

30+
Media sources
Moroccan + African
8
AI engines
ChatGPT, Perplexity, Gemini...
3
Languages
FR · AR · EN
60s
Crawl interval
continuous, 24/7
AI engine share — what your customers actually use
ChatGPT
68%
FR/AR/EN
Perplexity
14%
FR/EN
Google AI Overviews
9%
FR
Gemini
5%
FR/EN
Claude
2%
FR/EN
Copilot
1%
FR/EN
Claude
1%
FR/EN
Grok
0%
FR/EN
Ready to start?

See what the world says about you — for free.

We will run our full pipeline on your brand for 7 days and send you a sample report. No commitment, no credit card.

Get your free auditSee pricing