Enhancing Career Skills Management with AI: Gladeo's GPT-4 Integration


Gladeo is a career navigation platform that uses storytelling and videos to help put people on a successful career development path. Gladeo used to spend a lot of time maintaining a small skills database. We were able to massively reduce maintenance time using GPT-4, greatly increase the number of skills in the database, and include a manual audit tool to refine and verify results.



  • Managed Services, DevOps, and Automation


The Gladeo platform has information on thousands of careers. One attribute of a career is the associated skills. Managing the skills for each career could be overwhelming—especially because the skills for a career can change over time. 

Gladeo wanted to create a large database of skills that are needed for each particular career. They were interested in leveraging Open AI Large Language Model technology to generate a current, exhaustive list of skills instead of manually researching and adding them periodically. 



We built an application that automatically queries OpenAI’s GPT-4 to generate skills for a specific job. The request prompts were carefully crafted so that we received the most useful response. Still, it’s important to control what is displayed on the site, so we built administrative tooling into the Drupal administration interface so that Gladeo personnel can review and edit the skills that are autopopulated from the application. 

Generative AI tools don’t send perfect information. That’s why an administration tool is essential so that Gladeo administrators can modify the results when needed. After the skills database is populated, the skills list can be easily edited.



Challenges And Findings

One challenge is having the application send the right prompt to GPT-4. You have to finesse what you are asking GPT-4 so you can receive the type of answer you are looking for. It takes some trial and error as you receive back insufficient or sometimes inappropriate results. A good prompt is clear and provides important context and constraints. The better you guide GPT-4, the better results you will achieve.

We have several findings in our quest to write the perfect prompt:

  • We have to specifically request that GPT-4 does not respond with any unethical or criminal skills.

  • We could ask GPT-4 to label and separate “soft skills” from “hard skills.”

  • GPT-4 could add separate information about related certifications.

  • GPT-4 can assign a “percentage of importance” to each skill, so we can rank them.

  • We can get the response in exactly the format we desired.



By leveraging GPT-4, Gladeo was able to establish a very large skills database with minimal effort. Administrators have the ability to easily edit the results before they appear on the site. And because there may be changes to a career’s required skills over time, administrators can periodically run the application to get updated results.

Gladeo resultt.png