“A system is never the sum of its parts, it's the product of their interaction” - Russell Ackoff

As we look toward integrating AI tools like ChatGPT in the workplace, it’s never just a simple change. Like any workplace process, Learning & Development (L&D) workflows have interdependencies between tasks and key individuals with varying roles and responsibilities.

While introducing AI in the process can streamline tasks, it also creates the need for new tasks and steps (like content validation). In this respect, AI won’t replace instructional designers, training managers, or subject matter experts in the process, but it does create the need to adapt and adjust roles.

Mapping out your AI enhanced L&D workflow can help you better understand how roles will shift, identify performance support your employees may need, find gaps before implementation, and more.

Example High Level AI Enhanced L&D Workflow

This is an example of what an AI enhanced L&D workflow for creating a compliance training course could look like. In this scenario, previously, the instructional designer (ID) would have had a much larger role in outline and course content development. As this new workflow indicates, AI tools have largely taken the lead role in these two tasks and the ID role shifts toward prompt engineering and content review.

The dynamic between the ID and subject matter expert (SME), the compliance department, in this scenario also shifts. While traditionally the ID/SME partnership requires spending significant time communicating via interviews, emails, and more, an AI enhanced workflow changes this relationship. The SME no longer needs to be the primary source for raw material. Instead, their role shifts from providing knowledge to reviewing for accuracy. This allows them to focus less on generic knowledge sharing and more on how it specifically applies to your learners.

Although this type of rapid development and iteration can streamline the process, you’ll also want to consider your AI tool limitations. For example, during course creation, if you have limited usage of a particular tool used the process, you’ll want to bring in your SME for a review during content creation before it moves to that stage. By taking a hybrid waterfall/agile approach to content development you can better tailor your process to your specific situation.

Example Task Level AI Optimized L&D Workflow

This is an example of what a specific part of a process looks like in an AI optimized L&D workflow. This particular workflow focuses specifically on creating course content (the step just after outlining and before building the course).

In this scenario, the AI tools are doing much of the heavy lifting. But this doesn’t fully replace the ID role in the process. Throughout the workflow, the ID is continuously reviewing outputs, determining if they need adjusted, and making manual updates when needed.

As new tools and models emerge (especially OpenAI’s much anticipated GPT-4), the number of different AI tools needed in the workflow will likely decrease.

Workflow Inputs & Job Aids

In addition to building out an AI enhanced workflow, identifying key inputs and performance supports, or job aids, can help you more effectively implement your process. Examples of inputs could include your organization’s AI policy, your use cases, and your needs assessment. Some job aids to consider are a validation checklist for SMEs, a first-pass validation checklist for the ID to use before handoff, a data security checklist for building prompts, prompt templates, and more.