Blog
Creating personalized briefs with AI: a teaching case study
In December 2025, I started teaching at CIFACOM with third-year Bachelor Graphic Motion Designer students, on a module dedicated to video explainers.
The project was fairly classic in shape: design and produce a short explanatory video in 2D motion design, between 60 and 90 seconds, with voice-over, music and sound design. But I wanted to avoid a common trap in teaching exercises: the overly generic brief.
A single topic for a whole class is easy to manage and evaluate. But it often produces very similar answers. Offering three or four topics is already better, but it is still limited. With only a few options, the chances that every student finds a subject they genuinely care about are pretty low.
This is the context in which I started using Skills, shortly after Anthropic launched them.
A skill is not just a prompt
In the Claude ecosystem, a skill is a way to give an AI assistant a specialized capability.
It is not just a prompt you copy and paste to get a one-off result. It is closer to a reusable mini workflow: a frame, rules, steps, examples, sometimes reference files.
The point is to turn a general-purpose assistant into a more precise tool for a given task.
In my case, the task was this: generate personalized video explainer briefs that were realistic, pedagogically coherent, and varied enough for each student to work on a subject they could actually engage with.
The full skill is available here:
Read the brief-generator-videoexplainer skill
The pedagogical frame of the module
The module was part of the Bachelor Graphic Motion Designer program, aligned with a level 6 RNCP certification. It was organized around two main blocks.
The first block focused on concept development: analyzing the brief, developing the narrative concept, defining the visual world, creating the storyboard, producing style frames and an animatic, then presenting the concept.
The second block focused on production: creating 2D assets, animating in After Effects, organizing files, optimizing the render, documenting the process and preparing the jury dossier.
The final deliverable was a short explanatory video in 2D motion design, with complete image and sound.
That structure came with several strong constraints:
- a precise duration, between 60 and 90 seconds
- 2D production only
- a voice-over or placeholder voice
- music and sound design
- clear documentation of the process
- professional file organization
- the ability to justify narrative, visual and technical choices
The personalized brief therefore had to remain compatible with these constraints. The goal was not to let AI freely invent an attractive topic, but to adapt an already defined pedagogical frame.
The problem with generic briefs
In a motion design class, the choice of brief has a direct impact on engagement.
If the subject is too abstract, too academic or too distant from the students’ interests, they can complete the exercise correctly without truly investing in it. They respond to an assignment, but they do not really take ownership of the project.
On the other hand, when a student works on a subject that speaks to them, the dynamic changes. They look for more references, defend their choices better, accept constraints more easily and understand production trade-offs more concretely.
The challenge was to build a system that allowed for:
- real personalization of topics
- fairness in evaluation
- a comparable level of difficulty
- professional realism
- a direct connection to the targeted RNCP skills
This is exactly the kind of problem where a skill can become useful.
How the skill worked
The skill asked the student four questions.
- Which theme did you choose?
- Which specific subject do you want to cover?
- Which type of client or commissioner did you choose?
- What is the specific identity of that commissioner?
The themes were deliberately broad: historical figure, innovative product, history of an invention, artwork, pop star.
The commissioner types were inspired by credible professional contexts: ARTE, museum, Instagram influencer, YouTuber, company or startup.
From that combination, the skill generated an adapted brief.
For example:
- Alan Turing for an ARTE short program
- Monet’s Water Lilies for a mediation film in a museum
- E-ink for a YouTube segment
- A cleantech product for a startup fundraising video
- David Bowie for an Instagram format
The system was simple to use, but rich enough to produce very different briefs.
An ARTE brief does not sound like an Instagram DM
One important aspect of the skill was the simulation of the commissioner type.
An ARTE brief had to feel like an institutional specification document: broadcast context, editorial line, technical constraints, deliverables, contacts, validation process.
A museum had to express its request through cultural mediation: exhibition, visitor journey, learning objective, installation context, broad audience.
An Instagram influencer could send a much more informal message, sometimes vague, with references to a “vibe”, a duration constraint, an expectation of a finished piece and very few details about budget or validation.
A YouTuber would usually send something more structured but pragmatic: a segment to insert into a longer video, consistency with the channel, placement constraints, sometimes a sponsor relationship.
A startup would frame the request through business goals: pitch, sales, fundraising, trade show, value proposition, metrics, deadline.
Working on the form of the brief mattered. It helped students understand that professional requests do not all look the same. The project changes, but so does the way the client expresses the need.
Deliberately incomplete briefs
The skill was not supposed to generate perfect briefs.
On the contrary, it had to produce realistic briefs, which means sometimes incomplete briefs.
That was a central pedagogical choice.
In real life, a client does not always provide a clear narrative angle. They do not necessarily specify the expected level of vulgarization. They may forget to mention the validation process, budget, accessibility constraints, number of revisions or visual references.
These grey areas were deliberately built into the briefs.
The goal was to train the C1.1 skill from the concept block: analyzing the context of a client or commissioner request.
Students therefore had to learn to distinguish:
- what is explicitly requested
- what is implicit
- what is missing
- what should be asked to the client
- what can be formulated as a working hypothesis
This is a very practical skill. It prepares students better for professional work than a perfectly locked brief where all the framing intelligence has already been done for them.
Personalizing without breaking evaluation
The difficulty with personalization is fairness.
If every student has a different subject, how do you keep evaluation coherent?
The answer was not to personalize the assessed skills, but only the context in which they were applied.
All students worked toward the same objectives: analyzing a brief, developing a narrative concept, defining a visual world, producing a storyboard, creating style frames, building an animatic, then producing a final video with professional organization.
The evaluation grid remained common.
What changed was the world in which these skills were exercised.
One student might work for a museum, another for a YouTube channel, another for a startup. But they all had to demonstrate their ability to transform a request into a coherent motion design project.
To me, this is one of the most interesting uses of AI in teaching: allowing for more variety without losing the frame.
Two work fronts: narration and art direction
The module also relied on two parallel work fronts.
On one side, the narration front: text, structure, voice-over, timing, clarity of the message.
On the other side, the art direction front: moodboard, references, style frames, composition, palette, typography, visual world.
This organization was particularly useful for a short format. On a 60 to 90 second video, the text cannot be treated as a formality. It has to be read, recorded, timed, cut, rewritten and tested rhythmically.
At the same time, art direction needs to move forward early enough to avoid ending up with a strong text but no solid visual language.
The personalized briefs fed both fronts. The commissioner type influenced the narrative tone, but also the visual references, distribution format, density of information, pedagogical level and presentation style.
What AI actually brought to the process
AI did not replace the pedagogical work.
It mainly allowed that work to scale.
Without the skill, manually producing that many differentiated briefs would have been possible, but time-consuming. Each request would have needed to be written, each tone varied, each constraint adapted, each grey area designed, while keeping everything coherent with the schedule and assessed skills.
The skill made that logic systematic.
It did not decide the pedagogical frame. It applied it.
It did not replace the teacher’s judgment. It produced a first working material that I could review, adjust and integrate into the course.
That nuance matters. A poor frame only produces poor briefs faster. A good frame allows AI to generate useful variations.
What I observed with the students
The result was very encouraging.
Students took much stronger ownership of their topics. The projects were more varied, the references more personal, the discussions more concrete.
We were not just talking about “making a video”. We were talking about target audience, distribution context, client, level of explanation, technical constraints, art direction, planning and production dossiers.
The transition to the production block also showed the value of that concept phase. Students were not starting from an abstract exercise. They were turning a chosen, framed and defended concept into assets, animation, transitions and renders.
We will have to wait for the final jury to fully assess the impact on results, but several films have already surprised me by their level of finish and investment.
What I would do again, and what I would improve
I would clearly use this system again.
I would probably strengthen three points.
First, the brief questioning phase. The grey areas were present, but it would be useful to go further by explicitly asking students to formulate a list of questions for the commissioner before starting concept development.
Second, decision traceability. When a student fills in missing information with a hypothesis, they should document it as such: “The brief does not specify this point, so I am making the following assumption.”
Third, the comparison between the initial brief and the final project. It is an excellent exercise for understanding the gap between request, intention, production and result.
Why this use case matters to me
AI in education is often discussed through cheating, automation or content generation.
Those issues are real, but they are only part of the question.
In this experience, AI was not used to do the work instead of the students. It was used to create better working situations for them.
More personalized, more realistic, more demanding situations.
That is the point I find interesting.
AI becomes useful in teaching when it helps design the learning environment more precisely. Not when it artificially simplifies the task, but when it helps connect students with a complexity closer to real work.
For me, this is one of the most promising uses of agents and skills: building finer, more adaptable pedagogical setups without giving up structure or standards.
Read the skill
For anyone curious, I have published the skill used to generate these briefs.
It is not presented as a perfect model or universal solution. It is a working document, designed for a specific module, with its own constraints, goals and limits.
But that is precisely what makes it interesting: it shows how an AI use case can be anchored in a real teaching situation.