Select the knowledge the AI use case needs
A corpus suited to the mission and to the audience.
The use cases
AI use cases start from a need, not from a tool.
One base feeds several AI use cases. Choose your starting point: teams, customer relations, content or automation.
Seven AI use cases, one knowledge foundation.
7 use cases
Équipes & RH
Internal assistant
Find a procedure, prepare a meeting or support a new joiner.
Explorer cet usageSites web & accueilWebsite chatbot
Help your visitors find an answer and move on to the next step.
Explorer cet usageSupport & relation clientCustomer service & support
Help advisers find the right procedure and draft an answer.
Explorer cet usageMétiers & automatisationAI agents
Give an agent a precise mission, with the right knowledge and clear limits.
Explorer cet usageMarketing & communicationContent & campaigns
Prepare campaigns and posts from your expertise and your tone.
Explorer cet usageSEO & équipes webSEO analysis
Explore a corpus of pages, spot the subjects to deepen and prepare your briefs.
Explorer cet usageDocumentation & expertiseKnowledge wiki
Explore an approach that connects, summarises and makes your corpus easier to navigate.
Explorer cet usageNo use case matches this filter.
What they share
AI use cases change. The knowledge foundation stays.
Choose the rules
One model, coherent access rights and clear limits.
Plug the use case in
Inside Narratheque, or through an integration to be looked at.
Frequently asked questions
What you want to know about AI use cases.
Should we pick just one use case to start?
Yes, and as narrow as possible. One observable task, one identified audience, a small corpus: that is what lets you judge the result. The other use cases plug into the same base afterwards.
Can one base serve several use cases?
That is the principle. The same corpus can feed an internal assistant and a public chatbot, provided you decide what is visible to whom. Separation comes from access rights, not from duplication.
How do we know whether a use case is realistic?
Ask three questions: does the answer exist somewhere in your documents, could a person give it from those documents, and would you be able to tell whether the answer produced is a good one?
What if the sources change often?
You have to decide who updates them and how often. A corpus left as it is ages, and the answers age with it: that is the main thing to watch over time.
Let us start simply
Other AI use cases in mind?
Let us describe the sources, the task and the application to connect.
