Competency modeling involves identifying and documenting the skills and abilities necessary to be successful in a given role or function. Much of this process is iterative between the talent professional and subject matter experts (SMEs) to land on that perfect blend of what it means to be an ideal, successful occupant of the role being modeled. While it’s valuable to have these clear definitions for necessary competencies, it can be challenging to hash out definitions, especially writing behavioral indicators, which outline objective and observable behaviors representative of a given competency.

So, what does this have to do with artificial intelligence (AI)? Generative AI programs like ChatGPT, Google Gemini and others allow users to quickly generate content based on prompts. The iterative nature of competency modeling, coupled with what would otherwise be a cumbersome task of ideating competency definitions and behavioral indicators and committing them to writing, makes it an optimal candidate for generative AI capabilities. Rather than turning to already established, generic competency libraries, AI allows for the potential to easily and rapidly draft competency models tailored to the unique needs of your organization and your competency modeling strategy.

This article will outline several best practices and prompt engineering tips for using generative AI to develop competency models.

1. Make sure you both agree on the definition of the role.

If you already have a job description available, and if your organization’s AI policies allow, it can help get you and the AI on the same page to feed that description in at the start of your competency project. This way the AI understands the unique context and demands of the role it’s being asked to model.

If you don’t have an official job description, don’t fret, the AI can help with that too.

Example Prompt: Provide a job description for the role of bank teller.

Result: The AI gave me a robust job description, including a list of responsibilities with things like, “Handle and resolve customer concerns, complaints, or discrepancies promptly and effectively” and “Stay informed about changes in bank policies, procedures, and services to accurately inform customers.”

At this point, you can customize, tell the AI what aligns to the role you’re modeling. This could be a good point to run it by your SMEs and ask them to verify or edit as they see fit. Keep in mind, you can always customize your prompts as much as you want, for example:

Customized Prompt: Provide a job description for the role of an entry level bank teller working for a large national bank, out of a branch located in Manhattan.

Once you and the AI both agree on the job role, it’s time to move on to identifying competencies. Keep in mind that for it to remember what you’ve talked about, you’ll have to work within a single conversation session, or be prepared to “remind” it with a saved version of your agreed upon job description.

2. Ask for more competencies than you need and scale back with SME input.

Example Prompt: Based on this job description, identify the 15 most important competencies for this role.

Result: As requested, the AI gives me back 15 relevant competencies, such as “transaction accuracy,” “professionalism” and “adaptability to technology” and some short definitions to go along with them. At this point, once again get your SMEs to help you narrow it down. You can even put all 15 in a survey and ask them to rank them or pick their top five (be sure to leave space for them to recommend additional competencies as well). This way, you’re only taking a bit of your SMEs’ time, and you have enough information to keep going.

Once your SMEs have helped you narrow down the most important competencies, it’s time to dive deeper into defining what the competency means, and what it looks like.

3. Let the AI draft and customize your definitions and behavioral indicators, but give it conditions.

Here, I reminded the AI to keep the same job role in mind, gave it directions for how long to make the definition, explained the levels of behavioral indicators I wanted and quickly gave it some best practices for behavioral indicators.

Example Prompt: Keeping the same job role in mind, for the competency of “customer service excellence” provide a three-sentence definition and three levels of behavioral indicators (low, medium and high proficiency). These behavioral indicators should be observable and consistent in structure to allow for ease of objective rating for potential assessment.

Result

This is the AI’s exact response to the prompt above.

Competency: Customer Service Excellence

Definition: Customer Service Excellence involves consistently delivering high-quality service to meet and exceed customer expectations. It encompasses effective communication, anticipating customer needs, and providing solutions with a positive and professional demeanor.

Behavioral Indicators:

Low Proficiency:

• Displays limited engagement and responsiveness when assisting customers.
• Responds to inquiries with minimal effort, lacking personalized and proactive communication.
• Rarely seeks opportunities to enhance the customer experience or address concerns.

Medium Proficiency:

• Demonstrates satisfactory customer service, addressing inquiries and fulfilling transactions with a standard level of politeness.
• Provides accurate information when prompted but may not consistently anticipate customer needs.
• Occasionally seeks opportunities to enhance the customer experience.

High Proficiency:

• Consistently goes above and beyond to create a positive and personalized customer experience.
• Proactively anticipates customer needs, offering tailored solutions and additional assistance.
• Exhibits exceptional communication skills, actively engaging customers with a friendly and professional demeanor, ensuring a memorable and satisfying interaction.

Here, we have a workable draft that can be brought straight to an SME for feedback, after only prompting the AI with a few questions and very minimal required effort from our SMEs. Keep in mind, you’ll want to edit the AI output based on what makes sense for your organization and the needs of your particular competency modeling initiative. Whatever you send to your SMEs for feedback, you want to be refined and nearly ready to roll out according to your company’s standards. If you give them something completely out of touch to their reality, they’re likely to lose faith in your ability, and may lose faith in the entire project.

You can also ask the AI to customize existing competency definitions to different roles or needs. So, if you find a generic competency definition, you can feed that into the AI and ask it to customize based on your role. Take our example above, it’s not particularly unique to the role of a bank teller, so let’s ask the AI to try again.

Example prompt: Customize this competency definition and behavioral indicators to the unique needs of an entry level bank teller working for a large national bank, out of a branch located in Manhattan with bank-specific examples.

Result

The AI was able to customize, for example the low proficiency behavioral indicators now read as:

Low Proficiency:

• Demonstrates limited engagement and responsiveness when assisting customers, occasionally appearing disinterested in their financial needs.
• Responds to customer inquiries with minimal effort, lacking personalized and proactive communication about our various banking services.
• Rarely takes the initiative to seek opportunities to enhance the customer experience or address concerns specific to our Manhattan clientele.

Even though we had originally asked the AI to keep the role in mind, asking again and specifying that we wanted “bank-related” examples allowed us to combat what some have deemed AI laziness. More on that below.

4. Ask for bite sized information to combat AI’s “laziness”.

At this point, you could ask the AI for a definition and behavioral indicator for a whole list of competencies at one time:

Example Prompt: Keeping the same job role in mind, for the competencies of “customer service excellence,” “transaction accuracy,” “professionalism,” “adaptability to technology” and “knowledge of bank procedures,” provide a three-sentence definition and three levels of behavioral indicators (low, medium and high proficiency). These behavioral indicators should be observable and consistent in structure to allow for ease of objective rating for potential assessment.

I do not recommend this strategy because AI can get lazy (i.e., AI is programmed for efficiency that could result in less detailed responses the longer the requested output gets), and I find I get better results when I ask for what I need for each competency individually.

Finally, the last and most important tip:

5. Get SME input as often as possible, and never rely on AI results for your final product.

The AI may be out of touch and cannot fully understand the implied meaning of the words it generates, so your SMEs will be critical in shaping and validating the content provided by the AI.

Along with being customized to your unique organizational context, the best competency model for your needs will depend on how you intend to use the model. This will also rely on SME input. For instance, they can use their industry expertise and organizational knowledge to help you narrow down and identify competencies based on anticipated future needs, which can help support upskilling strategies by identifying and addressing competency gaps before they become a challenge. In another example, if you’re using competency models to support career paths, your SMEs can help you understand how different roles fit together within the organization and, thus, how their competency models should or should not relate or feed into one another.

The Training Industry Courses Competency Modeling Certificate is ideal for learning leaders seeking a process for developing effective custom competency models for their organization. Click here to learn more.