Modeling Consumer GenAI for Enterprise L&D Applications: Demonstration of typical use cases

Artificial intelligence (AI) and its applications are not novel developments. They have been in existence for several decades. However, with the advent of generative artificial intelligence (GenAI), such as ChatGPT, Google Gemini, Microsoft Co-pilot, and Claude, among others, understanding human language as a mode of interaction has become a reality. ChatGPT has chartered a revolutionary path with its hyper-realistic level of natural language processing, far beyond rule-based AI systems like IBM Watson. In addition to the other GenAI tools previously mentioned, ChatGPT can make sense of the context in which questions are asked, allowing for a natural conversation. This capability enables users to easily interact with GenAI using human language.

GenAI and Learning Processes

Given that ChatGPT and other GenAIs are content-based tools, they have great applications in learning and development (L&D) and training development areas. Creating a training course using traditional instructional design processes can be quite time-consuming. On the contrary, using GenAI can be a massive time-saver for course designers and educators. It quickly creates the structure and content for a course, as well as assessments and quizzes. Several solopreneurs, consultants, and influencers have shown early examples of using the free or paid consumer versions of ChatGPT (Model 3.5 or 4.0) to quickly create their course content for small-scale coaching or training projects.

However, applying GenAI to enterprise learning is a different ball game. Since 2023, there have been several powerful use cases of GenAI in the L&D space. Some of these are as follows:

• AI-driven instructional material development [1]
• AI-driven training course content development [1]
• AI-driven training video creation [2]
• Data analytics and predictive models [3, 4]
• 2D-3D visualization and animation [5]
• AI Augmentation for AR/VR [6, 7]
• AI-personalized learning path [8]
• Training, course management, and scheduling
• AI performance supports

This article focuses on the use case of GenAI for the first three items: AI-driven instructional material development, training course content development, and training video creation. L&D professionals can remodel their current enterprise learning and training development process by leveraging the power of ChatGPT (or any other GenAI tool) to efficiently create standard elements in a typical learning program. L&D practitioners can extend GenAI to a range of L&D deliverables while significantly reducing human effort, reducing cost, and accelerating the overall training program design process.  

Showcasing Enterprise L&D with GenAI

Let’s say you are in charge of designing learning programs for employees. Part of that challenge is to analyze the training needs, availability, and state of current content. This also includes creating course scope and outline and developing online or offline training material such as presentation slides, assessments, and videos. Moreover, accomplishing all of these tasks takes time.

However, for the sake of simplicity, I describe how you can accomplish each element of typical instructional design, development, and delivery using the free consumer version of ChatGPT (ChatGPT 3.5) and then scale up by implementing enterprise AI tools.

For this demonstration, I used ChatGPT 3.5 to build a short in-person training course to teach learners how to understand and apply Microsoft Excel’s new function, XLOOKUP, targeted at junior professionals who already know how to use the previous common function, VLOOKUP.

In the following sections, I demonstrate how I trained the ChatGPT model with a custom knowledge base and then prompted it step-by-step to build all the training course components, training content, lesson plans, course outlines, presentation slides, assessments, and videos, as part of the overall training development.

Step 1. Analyzing Training Gaps and Needs

Typical Enterprise Process

Typically, training specialists look at skills required for a particular job and then review performance data to identify gaps in the current training structure and content [9]. Based on this analysis, they determine certain performance or skill gaps and new training needs [10]. Generally, this leads to determining skills or competencies that must be taught in a new or revised training program. This assessment is usually done manually through back-and-forth conversations with experts by making qualitative references from various sources or inputs. The accuracy of this process is as good as that of the people involved in conducting the gap analysis, besides being highly effort-intensive and subjective. This is where GenAI tools come in, as they can perform the analysis objectively and within a few minutes. 

Extending GenAI to Achieve This Goal

For this demonstration, I scoped my goal only to perform training content gap analysis, which involved analyzing existing training content for potential gaps and ascertaining what additional content should be developed to deliver the identified skills.

I found a slide set on XLOOPUP that was written as a mini-training presentation. I referred to this as "available training content.” I used the AI tools ChatDoc and Humata, both of which have the ability to read documents in various formats, such as Word or PowerPoint. I prompted it to analyze the “available training content.” I provided it with intended learning objectives and asked it to report any information gaps that must be covered in the training course (see Figure 1).

Prompt: <What information should I add to make this document more comprehensive?>

Figure 1. ChatDoc input and output to analyze content gaps.

[click to enlarge]

GenAI suggested adding a more detailed explanation of certain topics. These gaps or suggested topics can then be used to develop complete training material.

Implications for L&D

Training content gap analysis is typically a small step of the bigger training need analysis exercise. This exercise is becoming more data-driven, given that large-scale LMS systems now allow a range of learning analytics. In this case, several new GenAI tools (Popai, MindGrasp, ChatDoc, Viable, DocLime, etc.) can be handy in analyzing a given set of documents, such as presentation slides, vendor documents, procedures, field data, student performance data, student feedback, reviewer comments, and other sources. Based on that, GenAI can identify patterns of skill or performance gaps against the specified learning goals. When automated, such a process can reduce a massive amount of human time.

2. Creating a Training Outline and Structure

Typical Enterprise Process

Creating a course outline, modular structure, and module outline is usually the first step in any corporate course creation [11]. Typically, this process involves thoroughly reviewing relevant and available documentation (which may include formal standard operating procedures), field notes, and training gap analysis with stakeholders and consulting them. Like other training development processes, this also requires human intervention and multiple discussions, reviews, and approvals from stakeholders.

Extending GenAI to Achieve This Goal

GenAI can be utilized to perform that task by leveraging available documentation to create a training course that includes hands-on activities, case examples, quizzes, and assessments. To demonstrate this, I prompted ChatGPT to assume the role of expert course designer to create a course description and outline for the XLOOKUP course. When you provide the GenAI tool with context, the outputs are far more accurate and specific.

Prompt: <You are an expert on course design and you are also an expert on Excel. Your job is to create a training course for professionals who have basic understanding of excel. The course is to teach them XLOOKUP function. The course should have hands-on activities as well as case examples to practice the function.>

For input documentation, I instructed it to use only the information or content fetched from two XLOOKUP procedures from two different URLs. These two URLs outline XLOOKUP function procedures that can act as standard operating procedure in this case, based on which training content will be developed. (If you wish to reproduce the outputs, here are two sources from CareerFoundry and Ablebits. Both procedures are completely distinct, written by different authors, and provide different explanations of XLOOKUP.)

Prompt: <Create a course description, outline, and lesson plan for a course on XLOOKUP. Use the information contained in procedures at URLs X and Y to create the outline and content.>

Within seconds, the AI created a high-level module and sub-bullets for the course, including topics such as advanced XLOOUP operations, best practices, and case studies.

Then, I prompted it to add more details (see Figure 2) about each bullet point to refine the outline further and also create a lesson plan (see Figure 3).

Figure 2. ChatGPT output to create a training outline.

[click to enlarge]

Figure 3. ChatGPT input and output to create a detailed training outline.

[click to enlarge]

Implications for L&D

GenAI can be instructed to use a specific set of materials and quickly generate a module outline and a lesson plan. It can create a draft of the training course in a matter of minutes, which can then be refined further by the instructional designer. While the AI’s output may need to be revised by a human, the speed with which it creates this intelligence is impressive and could potentially be used to expedite the course creation process.

3. Developing Instructor’s Presentation Slides 

Typical Enterprise Process

Most training courses rely on instructor-driven presentation slides, whether in-person or remote training classes, as the primary delivery mechanism [12]. Presentation slides are facilitation tools that can be used to create storyboards for instructor-led videos. Even if a course is self-learning-based, presentation slides are prepared in some form, whether as a script or a storyboard. Given that GenAI tools are content-driven, we believe they hold great promise for swiftly creating the required presentation material.

Extending GenAI to Achieve This Goal

To demonstrate this feasibility, I prompted ChatGPT to develop an outline of slides by feeding it the course outline I created previously.

Prompt: <I would like you to create an outline for the presentation slides that the instructor can use to teach the above course. Limit to 10 slides maximum. Fill out enough details for bullets and sub-bullets.>

ChatGPT generated an outline by automatically organizing the content in the form of slides developed in a logical progression. However, it was a little thin in detail.

Then, I prompted it to expand the bullets into detailed content (see Figure 4). While the slides may not be the final ones for the training course, they can act as an accelerated head start for instructional designers to populate and correct the content depending on the program’s needs. 

Figure 4. ChatGPT input and output to create a lesson plan.

[click to enlarge]

Figure 5. ChatGPT input and output to create a detailed outline of presentation slides.

[click to enlarge]

From there, I pasted the ChatGPT-generated outline into another AI tool called Gamma AI (see Figure 5). Gamma functions as an automated AI tool that converts minimal text or even high-level ideas into full presentations. It simply extracts instructions from the content you provide.

Figure 6. Gamma input to create slides from the basic outline.

[click to enlarge]

The Gamma tool created a full deck of slides instantly based on the context of each bullet point (see Figure 6). It also automatically selects appropriate images for the content and context and provides the slides in the most suitable layout to support the goal. The same function can now be done using ChatGPT 4.0 plugins or other AI apps like Tome.

Figure 7: Gamma AI output to create slides from the basic outline.

[click to enlarge]

Implications for L&D

GenAI tools can help eliminate manual work and swiftly create a draft representation of the presentation slides. The AI engine can design slides based on a specified skeleton of content. It follows a structure, is well organized, and has all the necessary details. You can force it to use contextual images from a specific repository. In most cases, it provides enough structure to make it easier for instructors to populate relevant information and quickly create engaging presentations.

4. Creating Problem-Solving Case Scenarios

Typical Enterprise Process

Most jobs now require higher-order problem-solving, technical, troubleshooting, and decision-making skills. L&D professionals greatly emphasize building scenario-based, thinking-intensive content that allows learners to practice on certain expectations [13]. Typically, scenarios are designed based on field events or canned scenarios provided by SMEs. It is a manual process conducted by instructional designers and SMEs, which can be quite time-consuming. However, GenAI tools can help us streamline the effort required and expedite the course development process.

Extending ChatGPT to Achieve This Goal

I prompted ChatGPT to create one case study for students to solve (see Figure 7).

Prompt: <Create one case study for students to solve. The case study should focus on fixing errors in the XLOOKUP function output. Create the case study in an interesting story format.>

ChatGPT generated a detailed story-based self-contained case study with necessary variables like situation, problem, constraints, and questions to test learners’ deeper understanding and help them troubleshoot common scenarios when using the XLOOKUP function (see Figures 8 and 9). 

Figure 8. ChatGPT input and output to create case study assignments.

[click to enlarge]

Figure 9. ChatGPT output to create case study assignments.

[click to enlarge]

Implications for L&D

Using GenAI tools, you can create powerful case study assignments by instructing the tool to focus on critical aspects of a challenge. You can instruct it to develop a story-based case study for students to solve. Moreover, if relevant materials for developing case study exercises are provided, GenAI can create well-structured case study assignments.

5. Create Learners’ Assessments and Assignments

Typical Enterprise Process

Most training courses include assessments, quizzes, or exams during and at the end of the training modules. Depending on the job, the nature of the assessment varies. The assessment can be in various formats, such as multiple choice, short answer, or essay type [14]. Here, you can leverage the power of ChatGPT to build systematic and highly targeted assessments for the training program.

Extending GenAI to Achieve This Goal

I gave ChatGPT a hypothetical context, stating that the course is targeted for two days. I prompted it to suggest and create an appropriate take-home assignment (see Figure 10). I also included a prompt for it to create a course-end assessment using the information provided for the course.

Prompt: <What ideal homework should I give the students if the course is spread over two days? Create 2 versions of thinking-intensive homework with challenging assignments. Generate an example of a homework assignment. Generate a 10-question assessment to test students at the end of the course.>

Figure 10. ChatGPT input and output to create learners’ homework assignments.

[click to enlarge]

ChatGPT generated a list of appropriate questions aligned with the scope of the course and level (see Figure 11).

Figure 11. ChatGPT input and output to create learners’ assessments.

[click to enlarge]

Implications for L&D

With appropriate prompts and scoping instructions, GenAI can be customized to provide more challenging assignments. ChatGPT can be used to accelerate the process of creating student assignments and various types of student work and quizzes.

6. Creating Self-learning Content and Videos

Typical Enterprise Process

More and more training programs are now designed using a blended learning approach. Training material now relies heavily on self-paced content and videos [15]. However, creating self-paced material, especially videos of good quality, is very time-consuming, requiring a range of logistics from cameras, software, editing tools, and other hardware. With GenAI, the content can be created with minimal effort and reduced time.

Extending GenAI to Achieve This Goal

I used the presentation slide content from Gamma and imported it into another AI tool called Pictory (see Figure 12). This AI tool converted the presentation slides into a storyboard. It then automatically sequenced all the scenes and video clips contextually. It selects relevant stock footage based on the context and content provided. The enterprise-hosted GenAI tools allow this selection to be contained in corporate repositories only, thus producing highly contextual videos.

Figure 12. Pictory output to create self-paced videos.

[click to enlarge]

The created video can be edited for scripts, scenes, clips, and sequences. The video thus created is a faceless video that can be used as online training material.

I thought of making it more attractive in an instructor-style video. Normally, it would have required a considerable setup of greenscreen, cameras, and post-shooting editing, which is a massive investment in cost and time. However, I used another AI tool called D-ID, as shown in Figure 13, to create a self-learning video in which I used my picture to create a near-realistic replica of myself to narrate the content as a video without having to shoot a single video (see Figure 13).

Figure 13. D-ID input and output to create instructor-led automated videos.

[click to enlarge]

Implications for L&D

Once you add an inbuilt voiceover or the one from your instructor, you can produce a ready-to-deploy, self-paced learning video in no time. Such videos provide learners with a personalized learning experience, sparing them from dispassionate “robo-lessons.” Some AI tools can intelligently read the transcript and segment or even edit the video into multiple short videos, which can be used appropriately for various learning paths.

7. Deploying GenAI as a Trainer

Typical Enterprise Process

While some organizations have transitioned to remote learning for training, instructor resources are still required to conduct formal training [16]. Instructor resources are the most expensive resources in any training program. However, as you see in the previous step, GenAI has the potential to leverage this to deliver the experience of a real trainer in the form of self-learning videos or as chatbot, with which learners can have a human-like conversation.

Extending GenAI to Achieve This Goal

To demonstrate this, I prompted ChatGPT to act as a trainer and teach me the specified content as if I were a 10-year-old.

Prompt: <I would like you to play the role of trainer. Explain to me how to use the XLOOKUP function with examples. Explain as if I were a 10-year-old. Reply to my follow-up questions, if any. At the end, you should ask me questions to test my understanding.>

Figure 14. ChatGPT input and output for using it as a chatbot trainer.

[click to enlarge]

It provided me with a slim-down version of the professional content that was fed into it earlier; it then presented the content in an easy-to-understand manner (see Figure 14).

As a learner, I can ask questions, clarify, and expand answers in whatever order I feel is right to understand them thoroughly. Once I provided the context to tone down and simplify the answers for a 10-year-old, it continued to respond with that constraint. For instance, it explained my subsequent queries with examples that are appropriate for a 10-year-old.

Implications for L&D

GenAI tools can understand the context and provide information specific to users’ queries or target audiences. It can be used as a tutor to guide and encourage learners. GenAI can be pre-programmed to ask sense-making questions to adaptively understand a learner’s grasp of the topic. This use case holds big promises by combining GenAI, chatbot, adaptive learning paths, and self-learning videos. Using the latest advancements in chatbots, you can engage learners during the learning process in a bidirectional conversation as if each of them has a personalized teacher by their side. Such engaging and lively conversation ultimately facilitates the learning process. In addition, real adaptive learning can happen because GenAI tools will progressively determine the learner’s progress and grasp of content and then change the learning path based on the learner’s knowledge acquisition level.

8. Deploying GenAI as an Automated Assessor

Typical Enterprise Process

Assessment of learners is usually a resource-intensive task, regardless of whether you use auto-grading in LMS against a preset answer key or assign an instructor to provide human grading and qualitative feedback [17]. GenAI can make this assessment efficient, deeper, and comprehensive for the assessment questions and feedback.

Extending GenAI to Achieve This Goal

I prompted ChatGPT to test my understanding of the XLOOUP function by asking me some questions (see Figure 15).

Prompt: <I would like you to test my understanding as a teacher, one question at a time. Check the correctness of my response to each question on a scale of 10, with 10 being the highest. Then give me feedback on what I did wrong.>

Figure 15. ChatGPT input and output to act as an adaptive assessor.

[click to enlarge]

It asked me questions to test my understanding. It gave me the correct answer with feedback and explained how I could improve my understanding. The interaction is friendly and conversational, like with a tutor. Depending on how well I could engage in a conversation, ChatGPT could go deeper into personalized feedback.

Then, I took it to the next level by assigning ChatGPT the role of an assessor to grade my understanding of XLOOKUP with questions. Not only did it provide instant feedback, but it also graded my responses and suggested the correct answers (see Figures 16–18). 

Figure 16. ChatGPT input and output to act as an adaptive assessor.

[click to enlarge]

Figure 17. ChatGPT input and output to act as an adaptive assessor.

[click to enlarge]

Figure 18. ChatGPT input and output to act as an adaptive assessor.

[click to enlarge]

Implications for L&D

GenAI, when implemented as a chatbot with a pre-programmed sequence of questions, can be used to deliver training and assess understanding. Learners will receive in-depth interactive feedback that accurately pinpoints their areas for improvement. It could even offer targeted lessons to close that gap.

9. Deploying GenAI as a Support Resource or an Expert Mentor

Typical Enterprise Process

Training is not a complete solution. Learners also need support in the field, which may come through peer help or expert mentorship. Organizations normally employ performance support systems such as learning portals, communities of practice, and buddy systems. Nevertheless, GenAI has emerged as a great workflow-based performance support system.

Extending GenAI to Achieve This Goal

To demonstrate this feature, I assigned ChatGPT the role of an expert problem solver. Then, I prompted it for help with an error.

Prompt: <I need expert advice. When I use XLOOKUP in my worksheet, I get #Value error. How do I fix it?>

Figure 19. ChatGPT input and output to act as a support resource.

[click to enlarge]

ChatGPT suggested a method to check and correct (see Figure 19). If the first suggestion did not work, it provided me with an alternative action plan until the issue was resolved. Thus, ChatGPT served as a helpful buddy or mentor to guide me on-demand during the workflow.

Implications for L&D

If you have new learners who have not received advanced training on specific topics, you can implement a chatbot expert that can be programmed to deliver expert feedback or suggestions in the context of the user’s problem. It brings the know-how to the audience’s level.

Scaling Up Consumer GenAI to Enterprise L&D

While most of the steps are invariably used in any standard instructional design project, some of the steps may not be applicable to certain organizations, depending on the nature of the business. However, there is a logical progression of steps to eliminate duplication and minimize efforts in such a way that you could apply the output of the previous step seamlessly into the following step. Figure 20 summarizes the nine steps demonstrated above as a quick reference. L&D professionals, course designers, and trainers can use this tutorial to conduct their pilots before deciding on full-blown AI implementation.

Figure 20. Nine-step process to model GenAI for enterprise L&D applications.

[click to enlarge]

Keeping a Few Things in Mind

You can create all the e-learning content in minutes by integrating various AI tools intelligently and strategically. While validating and refining the content may take some time, the results can be impressive.

GenAI is considered a valuable tool for creating new content in an efficient and contextually relevant manner. While the process of using ChatGPT or other GenAI for L&D projects looks very promising, there are several considerations you need to be aware of in enterprise settings.

• Monitor the big picture.

Technology has yet to go a long way, though it is rapidly evolving. We are yet to see the full potential of GenAI. Therefore, L&D leaders should exercise caution and look at the complete picture before committing significant investments based on consumer GenAI potential.

• Be aware of risks.

While all the new AI tools are promising, several unaddressed risks exist, including information security, intellectual property leakage, plagiarism, and liability issues [18]. A key concern among industry professionals is who is responsible for AI-generated content. Such legal implications for AI-generated content are still under discussion. You need to exercise caution before relying heavily on GenAI.

• Be sensitive to the impact on jobs.

We have yet to see the full social and employment impact of GenAI and other AI tools. A massive shift is occurring where traditional roles such as documentation writers, course developers, e-learning trainers, administrators, and reporters are transitioning into AI versions of those roles. Some of these roles have been automated using AI. Therefore, such an impact on jobs cannot be ignored. As a result, the critical decision would be how far and how fast organizations want to move into GenAI implementation.

• Be aware of information and IP security.

Information security is a key concern before L&D leaders can try any pilot program. Everything appears to be on track if you pilot using the Microsoft ecosystem. Microsoft’s enterprise version of ChatGPT uses Azure cognitive search to search through millions of documents to index the most relevant information in response to a prompt or question [19]. Microsoft’s architecture ensures all organizational knowledge, documentation, videos, and designs remain within the organization’s private knowledge base and are inaccessible to the ChatGPT model.

Readers’ Learning Resources

Interested readers can access a free, self-paced online course containing 25 videos demonstrating each step described in this article. The course is available at https://get-there-faster.com/enterprise-ai.

References

About the Author

Dr Raman K Attri is an L&D thought leader and founder of GetThereFaster Academy, specializing in coaching futuristic Chief Learning Officers. His research focuses on strategies and systems to shorten the workforce’s time-to-proficiency in the era of AI and speed. A Fortune 500 Technical Learning Leader and recipient of the CLO of the Year award, he is an active researcher, author, and speaker who takes pride in spearheading the science of accelerated organizational learning, enabling organizations to stay ahead in the fast-paced business world. He has authored more than 50 books, besides several magazine and journal articles on leadership, L&D, and training.