Everyday AI: the fundamentals
You leave with
Five uses tested on your own documents, and a method to check what the machine produces.
- 2 days · 14 h
- Beginner
- Remote or on site
On request
Writing effective prompts
One day to go from a prompt that works once to a prompt that works every time, on high-volume tasks.
You leave with
A library of tested, documented prompts that can be shared within your department.
Course information
- Duration
- 1 day · 7 h
- Level
- Intermediate
- Format
- Remote
- Languages
- Français · English
- Time to enrolment
- Two weeks between the first exchange and the start of the course.
- Accessibility for people with disabilities
- Remote session. Materials and pace can be adapted on request: write to us before enrolling.
- Issued on completion
- Statement of completion issued at the end of the course, stating the objectives, the duration and the assessment results.
Detailed syllabus
- What changes between a successful demonstration and high-volume use
- Judge a prompt on its worst output, not its best
- The three most frequent causes of drift
- Impose a format: mandatory fields, closed lists, length
- Explicitly plan for the case where the model cannot answer
- Produce a result another tool can reuse directly
- Build a representative sample of your real cases
- Compare two versions of a prompt on the same sample
- Set the acceptable error rate according to what is at stake
- Document a prompt for someone who did not write it
- Organise a shared library within a team
- Spot drift after a few weeks of use
The day's reversal
A demonstration prompt is judged on its best output. A production prompt is judged on its worst output.
The whole day follows from this reversal. It is what sets it apart from an introduction.
Learning objectives
- Structure a prompt whose result is reproducible
- Constrain an output format that is usable without rework
- Build a set of test cases, edge cases included
- Document and share a prompt across a team
Who it is for
- People already comfortable with a conversational assistant
- AI leads of a department or a team
- Freelancers producing content at a regular volume
Prerequisites
- Regular use of an AI assistant for at least a few weeks
- No programming skills
Teaching methods and materials
- Alternating demonstrations and individual workshops
- Comparison of two versions of a prompt on the same set of cases
- Prompt library template provided and filled in during the session
- Eight participants at most, one trainer
Assessment methods
- Placement questionnaire at the start of the day
- Continuous assessment by observing the output
- Final prompt written by the participant and tested in front of the group
- Satisfaction questionnaire at the end and three months later
The other courses
Automating your tasks without coding
You leave with
An automation running on one of your processes, documented and recoverable by hand.
- 2 days · 14 h
- Intermediate
- Remote or on site
On request
Tell us about your project
Describe your situation and we will come back to you with the syllabus or the approach that fits.
- Address
- 200 rue de la Croix-Nivert, 75015 Paris
Form
Reply within one working day, no newsletter.