How to get highly customized AI output
Introduction
As an instructional designer, you know that the impact of the training you create is directly tied to how well your course materials are tailored to your audience and your organization’s specific training objectives. If you’ve tried to use AI to generate training content, you’ve likely experienced results that are generic or misaligned with your needs. You may have even given up on using AI for creating high quality learning content. But you shouldn’t.
While you should always treat anything generated by AI as a draft, the surprising truth is that AI‐generated drafts can be clearer, more focused, and more effective than the drafts you write from scratch—particularly if you work under tight deadlines. You just need to provide your AI model with the right custom data, guidelines, and instructions.
The secret is a technique I call context modules. It involves creating a modular library of concise, re‐usable documents you can include with your prompts to tailor the output for different scenarios.
Combining human and computer expertise
The guiding principle behind context modules is to allow expert instructional designers to do what they do best while taking advantage of AI to do what it does best. Context modules allow you to pass key parts of your individual instructional design expertise to your AI model so that it can do a better job creating customized content.
The importance of creating a human‐AI partnership cannot be overstated. AI cannot replace the finely honed instructional sensibilities, understanding of organizational context, empathetic awareness of audience needs, or nuanced understanding of a topic that comes with human experience.
On the other hand, AI excels at synthesizing vast amounts of information and tailoring summaries for particular perspectives or audiences. It has access to extensive knowledge that can fill gaps in your expertise. It can generate draft content at incredible speed. And its pattern‐matching abilities allow it to spin off variations of content for different delivery formats, audiences, or levels of understanding nearly instantly.
The context modules approach provides several key advantages:
- Increased visibility and control ‐ Curate and improve the information your AI model uses so you can get more tailored, effective, and repeatable results.
- Fine‐grained refinement ‐ Test and improve individual parts of your AI prompts and data to fine‐tune AI responses.
- Modular efficiency ‐ Build a library of shareable documents that dramatically speed future learning development.
- Quality assurance ‐ Catch errors and biases before they influence your course content.
The ins and outs of AI context windows
At the heart of the context module approach is the concept of an AI model’s context window. A context window is how much information an AI model can hold during a single conversation. You can think of it as available working memory. The context window holds your prompts, any data or documents you include with your prompts, and the AI model’s responses.
When you upload documents with your prompt, the document contents are treated exactly the same as if you’d typed them into the prompt yourself. This means context modules can include instructions for the AI model as well as custom data.
Today’s top models have context windows large enough to hold entire books. So you might be tempted just to dump raw data into your prompts. But this presents challenges:
- The more diverse data you ask an AI model to process at once, the greater the likelihood that the summarization process will miss something important or draw an erroneous conclusion.
- With large amounts of data, you have limited visibility into how the data might influence your AI output.
- Raw data is almost always messy and contains information that is outdated, conflicting, or just plain wrong.
When you ask AI to summarize custom datasets one at a time to create focused summaries, micro‐level data issues tend to wash out so they don’t affect your results. You get an easily understandable preview of what information the AI model will use, and you can review and refine the summaries to catch errors, change emphasis, or tailor content to your learning goals.
Technical tips
Choose an appropriate AI model
A high‐quality, standard AI model is generally all you need to create your context modules. This is because the tasks involved align perfectly with the core strengths of most modern language models, which include summarizing data, extracting patterns, and synthesizing single topics from related data.
However, when you generate the final course materials, the task becomes significantly more complex. You are asking the model to synthesize information from multiple, diverse context modules simultaneously—this is where a more advanced “reasoning model” excels. Advanced reasoning models are typically only available with paid AI plans.
Note that different models will provide very different results. So test a few with your subject matter before deciding. Or, do what I do, and provide the same prompts and documents to two different models (I use Gemini 2.5 Pro and Claude Sonnet 4). Then merge the results to pull the best from each model. For instance, one might excel at creative examples while the other is better at logical structure.
Use AI‐readable formats
Different AI models accept and create different data and document formats. When using AI to summarize custom data or create context module documents, it pays to use simple, universal document formats such as plain text (.txt) or Markdown (.md). If you have tabular or spreadsheet data, use the Comma Separated Values (.csv) format if possible.
While some models can read proprietary formats such as Microsoft Word files (.docx), Microsoft PowerPoint files (.ppt), or Adobe (.pdf) files, this requires the AI model to convert these formats before executing the rest of your prompt. And the conversion may not capture all of the detail in the file. For example, when a model converts a PDF, it often loses crucial formatting like headings, tables, and columns, flattening the content into a single, hard‐to‐interpret block of text.
Another advantage of universal document formats is that you don’t lose the ability to use the documents if you switch to another model. It’s a better practice to pre‐process source files that are in proprietary file formats to convert them to universal file types. Depending on your AI model, you may be able to get AI to assist with document conversion. If you want to incorporate your AI output into a context module, don’t forget to specify a universal file types when generating the content.
Manage your library of context modules
Your context modules will evolve as you get more experience using them. You may also spin off variations of a particular context module for different purposes. This makes it critical to use clear names and some form of versioning when storing them. If you are using context modules to get repeatable results across a team, carefully construct and publish a file repository structure and naming conventions or use a dedicated document management tool.
The five essential context modules
Although I’m using the example of creating learning materials, the same approach works for a variety of content generation use cases. Simply adjust the number and types of context modules to match your needs.
1. Audience description
Start here because you’ll reference your audience description when generating other context modules. The most useful audience descriptions often combine AI‐generated descriptions of typical audience attributes with organization‐specific audience data.
When using AI to generate a description, use a template to specify the audience attributes that are important for your situation. Also, reference your preferred format in your AI prompt. For instance, do you want AI to generate a general audience description, a learner persona, or a list of role‐based competencies? If you have an audience description format you already use, include it as a pattern for your AI model to follow.
Sample audience description template:
# Audience: [Role Title]
## Role Definition
[2‐3 sentences describing their primary responsibilities]
## Key Characteristics
‐ Experience level: [Junior/Mid‐level/Senior]
‐ Education background: [Typical degrees or certifications]
‐ Technical comfort: [Scale of 1‐5 with explanation]
‐ Learning preferences: [Visual, hands‐on, theoretical, etc.]
## Training Context
‐ Typical session length they can attend: [30 min/1 hour/half day]
‐ Device usage: [Desktop/mobile/both]
‐ Motivation level: [Required vs. voluntary participation]
## Success Metrics
[What does success look like for this audience?]
Source materials include: Role descriptions, job postings, survey data, or transcripts of audience interviews.
2. Subject matter information
If you are an instructional designer, there’s a high probability that getting access to subject matter experts (SMEs) is the bane of your life. Don’t despair—AI can help.
For universal topics, you can use AI to create subject‐matter explanations that you can use directly in a context module. However, even with universal topics, there is likely organizational context that should be factored in. For instance, if you’re creating training on conducting employee reviews, your organization likely has a specific management philosophy or performance rating guidelines. You can add this custom information to your AI prompt to get more tailored AI‐generated content.
For more organization‐specific topics such as product training, you’ll need to use AI to generate subject matter modules from your organization’s custom data.
Useful source materials include:
- Product documentation and knowledge‐base articles
- Organizational policies and operating procedures
- Competency models and learning taxonomies
- Interview transcripts with leaders or subject‐matter experts
- Existing course materials
Pro tip: Don’t try to process everything at once. Create focused summaries by topic area, then combine them when building courses. Try to limit individual source documents to 10,000 words where possible.
3. Course parameters
This is a document that specifies course details such as delivery format, duration, types of learning activities, and output format. Unlike other modules, it is typically written by an instructional designer rather than generated by AI. Include as many details as you can to reduce the amount of editing you have to do on generated content down the road.
Sample course parameters module
Topic: Effective Feedback Conversations for New Managers
Duration: 25 minutes
Format: Virtual Instructor‐Led Training (vILT)
Facilitator profile: Facilitators have an HR background and extensive experience delivering live courses.
Audience: Recently promoted managers with 0‐2 years of management experience.
Learning Activities: Include one interactive poll and one 5‐minute breakout room discussion.
Output Format Requested: Generate a detailed instructor’s guide with talking points, facilitator notes, and specific instructions for the poll and breakout activity.
4. Key learning principles
Out of the hundreds of learning theories available, you probably rely on a few core principles that guide your learning design. Referencing these when working with AI ensures content that’s both instructionally sound and consistent with your existing materials.
In most cases, referencing a well‐known learning theory by name is sufficient. But it doesn’t hurt to have AI generate a brief summary you can include in your context module.
Common learning theories to consider:
- Adult learning theory (andragogy)
- Bloom’s Taxonomy (include your desired knowledge level)
- Cognitive load theory for complex topics
- Social learning theory for skill development
- Microlearning principles for busy professionals
- Accessibility and inclusive design requirements
5. Writing style and brand voice
No one wants to read content that sounds like it was written by AI. A writing style module ensures brand consistency and saves editing time. You can create this module by giving your AI model some writing samples and asking it to create a bulleted list of style elements in your sample text.
Style elements to specify:
- Formality level: Research paper, professional journal, blog post, or conversational
- Tone: Authoritative, supportive, encouraging, or straightforward
- What to avoid: Marketing language, clichés, excessive jargon
- Grade level: Use Flesch‐Kincaid scale (typically 8–12 for professional audiences)
- Special requirements: Inclusive language, brand terminology, regulatory compliance
Putting it all together: A practical workflow
When it’s time to use your modules to create learning materials, you’ll build up your course one step at a time, editing the result of each step to incorporate your unique expertise.
Start by including all the relevant modules in your prompt.
Sample prompt for combining context modules
Attached are five documents:
‐ audience‐new_managers.txt
‐ subject_matter‐effective_feedback.txt
‐ course_parameters‐vILT_25min.txt
‐ learning_guidelines.txt
‐ writing_style‐learning_objectives.txt
Using the information and instructions provided in these documents, please generate learning objectives for the training session.
The typical course creation process mirrors what you would normally do:
- Create learning objectives.
- Turn the learning objectives into an outline.
- Build lessons or modules one at a time—the output could be text, scripts, instructors’ guides, etc.
- Create enhancements such as learning activities, job aids, or other resources.
- Get feedback from potential users and from AI on the complete course.
- Run a pilot of the course (optional).
- Use AI to spin off variations for different audiences or learner levels by using different context modules.
When generating learning content, remember that the context window contains all of the previous AI conversation, including any errors or corrections. If AI responses start to veer off‐track during a long conversation, it’s safer to move relevant details to a new conversation and start over. This gives the model a clean slate with only your refined context modules and latest instructions, ensuring a more focused output.
The context modules technique transforms AI from a generic content generator into a sophisticated instructional design partner. By investing time upfront to create focused, reusable documents that capture your expertise and organizational context, you’ll generate consistently better results while dramatically reducing the time needed for future course development.
Happy course building.
