Skip to main content
Book
§ Use cases

What we deliver, concretely.

A handful of representative engagements, recent and recurring. Every project is calibrated to its context - never standardized.

  • Finance

    01

    Industrialize budget forecasting

    Context

    Budget cycles relied on Excel files passed between contributors, with information loss and manual consolidation.

    Intervention

    A structured input layer, a shared forecasting model, and real-time consolidated reporting. The finance department takes back control of the cycle.

    Outcome

    Budget cycle cut by 40%. Full traceability of assumptions. Automated consolidation.

  • Leadership

    02

    Consolidated executive dashboard

    Context

    The executive committee arbitrated on numbers reconstructed by hand, sometimes contradicting each other across departments.

    Intervention

    Unified data architecture, shared reference data, executive dashboard plugged into operational sources, refreshed without human intervention.

    Outcome

    One number per indicator, always fresh. Faster trade-offs. No more debates about the data itself.

  • Operations

    03

    Domain copilot for teams

    Context

    Teams spent significant time searching procedures, precedents, and contractual information scattered across documents.

    Intervention

    Document copilot on a controlled corpus: controlled ingestion, traceable sources, sourced answers. No public model exposed.

    Outcome

    Significant operational time savings. Controlled answer quality. Confidentiality preserved.

  • Data

    04

    Rationalizing scattered tooling

    Context

    Historical stack of a dozen reporting and data tools, with redundancies and high licensing costs.

    Intervention

    Systematic audit, substitution matrix, decommissioning plan, migration to three solid building blocks.

    Outcome

    Simplified landscape, recurring costs slashed, teams refocused on value.

  • Documents

    05

    Document automation

    Context

    Hundreds of recurring documents (contracts, slips, minutes) handled manually, with error risk and heavy workload.

    Intervention

    Extraction and classification pipeline, embedded business controls, assisted generation, structured human review.

    Outcome

    Processing time reduced. Quality leveled out. The team refocuses on analysis and arbitration.

  • Governance

    06

    Internal AI framework

    Context

    Uncoordinated AI initiatives, compliance risks, no shared view on usage.

    Intervention

    AI charter, usage scopes, evaluation process, model and hosting governance.

    Outcome

    Readable framework, scoped initiatives, executive sponsor reassured, ground prepared for scale.

§ In production

Two real engagements. A recent AI for regional tourism, and a language-processing module built from scratch.

Screenshot of Intellencia Tourisme - a digital concierge for regional tourism.● In production · June 2026
Case 01 · Tourism · Regional network

A digital concierge for regional tourism.

Client · A hosts network · regional tourism

Context - The hosts network wanted a hyperlocal conversational agent able to inform visitors in French, without depending on a public model.

Intervention - Design of the territorial knowledge base, RAG on institutional sources, configurable conversational interface, documented deployment. We also built an Analytics Dashboard to track usage: volume of questions, recurring topics, geographic zones, resolution rate - essential instrumentation to keep the knowledge base alive.

Results - Concierge in production since June 2026, hyperlocal, customizable per host or territory, embeddable in 30 seconds. Hosted by Infomaniak, in Switzerland. Visitor questions processed by Mistral AI. Analytics dashboard live to steer content over time.

Visit the live chatbot
● Built in 2020-2021
Case 02 · R&D · Vertical web application

A custom NLP tokenizer for proprietary semantic matching.

Client · A software vendor · Software for construction professionals

Context - A vertical web application needed to associate items by meaning, without depending on public models - full confidentiality constraints on the data processed.

Intervention - Custom NLP tokenizer built from scratch, proprietary semantic matching algorithm, full-stack module integration into the existing application, tests on construction-domain corpus.

Results - A language-processing module built from scratch for construction-industry software. No dependency on a third-party model. Full confidentiality preserved.

§ And upstream

And ten years of data work on SAP, inside large companies. That's the rigor that makes AI reliable.

Manifeste · Intellencia

« Every project is calibrated to its context. Never standardized. »

- Stance · Intellencia

§ The summit

Thirty minutes to find out whether AI is worth it for you.

  1. 0–8 minYou tellWhat takes your team's time, what changes this year.
  2. 8–27 minWe measure it togetherHow often per month, how many minutes, which data is involved, who decides.
  3. 27–30 minA straight answerA dated next step if AI is worth it. If not, Julien tells you, and what would work instead.
Pick a time slot

Booking goes through Calendly, which sets its own cookies. The booking page is in French. Write to Julien · Reply within 48 hours, no Calendly.

Email

contact@intellencia.ai

LinkedIn

Julien Bédouret ↗

Office

Poitiers, France

SAP assignments

Julien's SAP profile

The 30-minute call

With Julien Bédouret · video call · free

  1. 0–8 minYou tellWhat takes your team's time, what changes this year.
  2. 8–27 minWe measure it togetherHow often per month, how many minutes, which data is involved, who decides.
  3. 27–30 minA straight answerA dated next step if AI is worth it. If not, Julien tells you, and what would work instead.

Booking goes through Calendly, which sets its own cookies. The booking page is in French.

Write to Julien · Reply within 48 hours, no Calendly.