AI has made it easy to turn procedures, manuals and internal expertise into learning content, the first step in turning company knowledge into employee skills. Generating a course, however, is not the same as making it relevant to the people who have to learn from it.
A new hire, an experienced technician and a team manager may all need to understand the same procedure. They start from different knowledge, carry different responsibilities and will apply it in different situations. A single generic version is likely to be too basic for some, too dense for others and disconnected from the daily work of most.
Personalising AI-generated learning starts with a structured description of the learner, not with the source document.
Why generating a course from a document is not enough
When AI receives only a document and a topic, its main input is the content itself. It can summarise, restructure and shorten, but it has no basis for deciding which examples will resonate, which concepts the audience already knows or which decisions the learners will actually face.
Changing the tone or shortening lessons adapts the form. Meaningful personalisation changes what is explained, how deeply, with which examples and in relation to which responsibilities. That requires a second input alongside the source material and the learning objective: a model of the audience.
Personalising AI-generated learning starts with a structured description of the learner, not with the source document.
What is a Learner Model?
A Learner Model is a structured description of the characteristics of a group of learners that matter for how learning content should be designed for them. In AI-assisted course creation, it tells the AI explicitly who the course is for, so that generation decisions are not based on the source document alone.
In corporate learning, job title and seniority are only a starting point. Two people in the same role can differ in experience, familiarity with the topic, confidence and the way they prefer information to be explained. A multidimensional model captures these differences in a form the AI can use.
Two kinds of personalisation: during course design and during learning
“Personalisation” is used for two different mechanisms, often without distinction. Separating them makes any claim easier to evaluate.
Personalisation during course design | Adaptation during learning | |
|---|---|---|
When it happens | Before delivery, while the course is generated | While each learner takes the course |
Who it targets | A cohort or group with a shared profile | The individual learner |
Main input | Source material, learning objective, audience profile | Individual answers, history and behaviour |
What changes | Explanations, examples, vocabulary, depth, balance of theory and practice | Content, sequence or reinforcement for each person |
The Learner Model described in this article works in the first column: it guides the generation of a course for a defined audience. Adapting content to each individual’s answers in real time is a different capability and should be assessed separately.
The five dimensions of a Learner Model
A structured Learner Model can describe an audience across five complementary dimensions. Each one informs a different set of design decisions.
1. Professional Profile
Variables: Seniority, Experience, Function, Decision Autonomy.
These place the content in relation to the learner’s responsibilities and the decisions they make. Content about a new operational process might emphasise execution for the people carrying it out, and coordination and escalation for the managers who oversee it.
2. Prior Knowledge
Variables: Knowledge Level, Topic Familiarity, Confidence.
These shape how a topic is introduced and developed. Newcomers need foundations and clear examples; experienced staff benefit more from changes, exceptions and edge cases.
Knowledge and confidence are not the same. A group may know a rule well yet feel unsure applying it, which calls for more practice rather than more explanation.
Knowledge and confidence are not the same.
3. Communication Preferences
Variables: Explanation Style, Examples, Tone, Vocabulary.
These change how information is communicated without changing the underlying knowledge or the learning objective. Some audiences need a concrete example first; others prefer a direct description of the process. Vocabulary and tone can be matched to the people who will read the lessons.
4. Cognitive Profile
Variables: Abstraction Level, Reasoning Style, Theory vs Practice, Decision Style.
These guide the balance between principles, concrete situations, exercises and decision-oriented examples. A team that executes a procedure may need realistic steps and scenarios; a team that designs or audits it may need the underlying logic and the alternatives.
These dimensions are parameters for content design. They are not fixed psychological classifications, and they do not assume that people belong to rigid “learning styles”.
5. Business Context
Variables: Industry, Business Priorities.
Industry determines terminology, examples and operational situations. Business priorities connect the content to what the organisation cares about: a company focused on safety, for instance, will want lessons that emphasise risk recognition and correct execution.
How a Learner Model shapes course generation in practice
In zick learn, course creation combines the AI assistant, Lia, with the judgement of a Learning Designer. The Learner Model enters the process alongside the source material and the learning objective.
Sources.
The Learning Designer provides the source material: documents, procedures, manuals, and audio or video content.
Objective and audience.
The Learning Designer defines the learning objective, the skills to develop and the audience profile across the five dimensions.
Analysis.
Lia breaks the sources into atomic content units and classifies each one as core, supporting or to be discarded, taking the objective and the Learner Model into account. The Learning Designer can inspect each classification and the reasoning behind it, and change it.
Generation.
Lia generates the course: lessons, assessment questions and explanations of the answers, shaped by the audience profile.
Review.
The Learning Designer edits lessons and questions, and can regenerate individual lessons to change their length, complexity or depth.
The result is a course designed for a specific cohort, with editorial control in the hands of the people responsible for the training. When two audiences need different treatments of the same knowledge, each can have its own profile and its own version of the course, generated from the same source material.
Example: one procedure, three audiences
A company introduces a revised procedure for handling customer complaints. There is one authoritative document and three groups to train.
New employees
need definitions, the reason behind each step and worked examples from start to finish.
Experienced employees
need to know what has changed, the exceptions and the situations where an old habit would now be a mistake.
Team leaders
need to understand how the procedure affects escalation, team coordination and the priorities they have to balance.
The knowledge is the same in all three cases. What changes is the starting point, the emphasis and the decisions each group must make. These are illustrations, not prescriptions: the right treatment depends on the actual audience profile, the objective and the source material.
What real programmes show about audience and context
zick learn’s customer programmes show how different the audiences of corporate learning can be:
Poste Italiane trained 800 distribution centre directors, a managerial population with specific responsibilities and decisions.
Alperia trained 800 field technicians in Trentino, working across the territory, with content in four languages.
Gruppo Frassati delivers training that fits blue-collar shifts, where access to learning must work around operational schedules.
A pharmaceutical compliance programme turned existing training into shorter, multilingual learning for an international workforce.
These cases show why audience and context matter. They do not isolate the effect of any single Learner Model dimension on learning outcomes.
Language is a separate matter from personalisation. zick learn can translate an entire course into more than 200 languages in one click, which makes the same content accessible to more people. Translation alone does not adapt the content to a different audience profile.
What zick learn’s Learner Model does today, and what it does not
Precision matters here, because the same word can describe very different capabilities.
Today, the Learner Model:
Uses the five dimensions described above to guide course generation for a cohort or group.
Works alongside the source material and the learning objective, under the control of a Learning Designer.
Builds on a learning methodology co-developed with Berkeley.
Today, it does not:
Adapt lesson content in real time to the answers of an individual learner.
Automatically diagnose individual skill gaps.
Trigger reinforcement automatically from individual results. Recalls are scheduled by the organisation.
Learner data is still available for analysis. The platform records correct and incorrect answers, assessment results, activity, completion and progress, and exports raw data in CSV, as the zick learn FAQ explains in more detail. Organisations can use these exports to find the topics where a group struggles, then revise the course or its audience profile. Integrating individual learner data into the Learner Model is an area of ongoing development.
Questions to ask when a platform claims to personalise learning
Our guide on how to choose an AI microlearning platform for skills development covers the full set of selection criteria, from delivery channels to analytics. When the claim being assessed is personalisation, four questions help separate tailoring at design time from adaptation during learning:
What does the AI know about the audience?
A structured profile across several dimensions, or a job title and a free-text description?
What does personalisation change?
Explanations, examples and depth, or only length and tone?
When does it happen?
During course design for a cohort, during learning for each individual, or both? Ask to see each claim demonstrated separately.
Who stays in control?
Can a Learning Designer see why the AI kept or discarded content, and change that decision?
Frequently asked questions
What is the difference between personalised and adaptive learning?
A useful distinction is this: personalised learning designs content around the profile of a learner or group, while adaptive learning changes content or sequence in response to an individual’s performance during the course. The two can be combined, but they are separate capabilities.
Is job role enough to personalise corporate learning?
No. Role is a useful starting point, but people in the same role can differ in prior knowledge, confidence, communication preferences and business context. These differences affect how content should be explained.
Does a Learner Model mean every employee gets a different course?
Not necessarily. In zick learn, the Learner Model describes a cohort and the course is generated for that group. Different groups can receive different versions generated from the same source material.
Does AI personalisation replace instructional design?
No. AI can speed up analysis and generation, but accuracy, relevance and suitability still need to be checked by people who understand the content and the audience.
Conclusion: personalisation starts with a model of the learner
AI can make it easier to transform company knowledge into learning content. But generating content efficiently and personalising it meaningfully are two different challenges.
Meaningful personalisation requires a structured understanding of the intended audience and a way to turn that understanding into content-generation decisions. A multidimensional Learner Model, combined with human review, provides that framework.
For organisations looking to turn existing know-how into practical employee skills, the key question is not simply how AI can generate more learning content. It is how AI can generate the right learning experience for the people who need to apply that knowledge at work.
Once a course fits its audience, the next step is creating skills in the flow of work: bringing the course to people where they already work and supporting the application of what they learn.