Use cases
Use cases
Twenty-seven client cases in energy, pharma, industry, higher education and software publishing.
27 cases
Energy · IndustrySearchable technical reports, with the source of every answer
Decades of reports in several languages become searchable. Every answer points to the document and the page.Agentic platforms
EnergyReports, notes and presentations drafted from internal documents
Working from the document base or uploaded files, the assistant produces a risk report, a summary note, a contract comparison or a presentation ready for further editing.Agentic platforms
Higher educationInternal assistant on the document base
Teams query the document base, share their standard questions and have the assistant draft their meeting notes.Azure · Gemini · Claude
All sectorsRouting pages before ingestion into an assistant
Each page of a PDF goes either to text extraction or to reading by a vision model, only when that is useful. Ingestion is faster, and costly reading is used only on the pages that need it.AI in your applications
Energy · IndustryTenders: agents write a first draft of the technical proposal
The agents read the tender documents, search past responses and write a first draft of the technical proposal. Experts review and correct instead of writing everything.Agentic platforms
EnergyRequirement-by-requirement analysis of bids received by a buyer
When the group runs its own tender, the tool extracts the requirements from the specification and then rates each bidder's response, requirement by requirement, with the page that justifies the verdict.Agentic platforms · Forward Deployed Engineers
EnergyComparing supplier contracts against a clause library
Buyers import their clause library, upload contracts from several suppliers and get a matrix comparing each contract clause by clause, with an editable executive summary and Excel and Word export.AI in your applications · Forward Deployed Engineers
EnergyBusiness travel spend analysis
The team uploads exports from its booking and expense tools; the tool derives data coverage, savings levers, a forecast, the carbon footprint and an action plan, with every figure traced back to its source rows.AI in your applications · Agentic platforms
EnergyManaging a portfolio of R&D programmes
The R&D department tracks its programmes, projects and deliverables in a single view, filterable by year, with budgets, value created and source documents, and can query the portfolio in natural language.AI in your applications · Agentic platforms
All sectorsSourced opportunity note for launching a new programme
An agent produces the opportunity note for a new accredited programme: job market, competition, skills needs, with every figure sourced. Teams start from a structured file instead of a blank page.Agentic platforms · AI & Cloud Strategy
PharmaMapping marketing authorisations across 36 countries
A single workbook of authorised products, country by country, with the manufacturing sites. Every data point links to the official page and the text extract that supports it.Forward Deployed Engineers
Higher educationCompliance checks on apprenticeship contracts
Each contract is pre-checked on four axes: duties against the official skills framework, statutory dates, pay, and funding body. Flagged cases are presented for review.Forward Deployed Engineers
All sectorsAutomatic data entry in your software, without modifying it
The agent navigates the software and fills in fields as a person would, with the user's permissions, without modifying the software.AI in your applications
Software vendorsAgent-driven regression testing
An agent drives the browser, replays the scenarios described in a spreadsheet and logs any discrepancies.Azure
All sectorsChecking a supplier master file against public registers
A supplier file is automatically compared with the official company registers. Each row comes back with a status: up to date, activity changed, establishment closed, company replaced, or not found.Forward Deployed Engineers
All sectorsMapping company contacts from shared mailboxes
Departments' generic mailboxes become a map of relationships: who talks to whom in each company, and on which topics. Each thread is classified by theme by a language model.Forward Deployed Engineers · AI in your applications
All sectorsNatural language questions on the data warehouse
Teams ask their questions in French and an agent queries the data warehouse itself through SQL tools.AI in your applications · Agentic platforms
IndustryMeasuring building occupancy
The agent consolidates schedules from several sites and calculates the occupancy rate of rooms and buildings.Azure
All sectorsSafety guardrails for an assistant
Dangerous requests are stopped before they reach the assistant, on lightweight infrastructure.AI & Cloud Strategy
All sectorsMulti-model gateway with routing by complexity
Each request is classified as simple, medium or complex, then sent to the leanest model capable of answering it. The organisation can change models without redoing the integration.Agentic platforms
All sectorsTracking AI costs project by project
Cloud and model costs are consolidated in euros per project, with hard budgets, rate limits and threshold alerts.AI & Cloud Strategy
All sectorsAI infrastructure audit and replacement platform
The existing infrastructure is surveyed resource by resource, with its risks and the line that weighs heaviest. A replacement platform hosted in France is built and tested end to end.AI & Cloud Strategy
All sectorsAI experimentation platform with a catalogue and unified access
All AI experiments go through a single portal, with enterprise authentication, a request form and usage tracking. A new use case starts from a shared foundation.Forward Deployed Engineers
All sectorsMeasuring adoption and analysing conversations
Administrators track adoption of the assistant by segment and receive an analysis of conversations: topics, pain points, missing content.AI in your applications
All sectorsMeasuring visibility in AI engines
For each page of your site, you know whether AI engines cite it, and which pages to rewrite first.Citead
EnergySolar power generation forecasting
Next-day solar power generation, forecast the day before and compared with measured output.Time-series forecasting
All sectorsTraining a developer to work with a coding agent
An experienced developer learns to work with a coding agent by building a library that estimates the carbon footprint of requests to language models.Training & maintenance
We name a client only with their written agreement. The budgets and detailed results of our engagements remain confidential.