In the rush to adopt generative AI, a lot of platforms (including SkillUp) have jumped headfirst into the deep end adding chatbots, surfacing quick answers and hoping automation will bridge complex gaps in education and employment. But when you work with job seekers who face overlapping challenges like limited networks, financial strain or unclear pathways, AI can’t just respond to questions. It also has to help people act.
After piloting our own AI tool, this realization now drives a major evolution in how SkillUp builds, tests and deploys multiple styles of AI tools to support our users — many of whom are people skilled through alternative routes (STARs).
What We’ve Learned So Far
Our AI journey began with SkillUp AI, a chatbot project supported by the Gates Foundation, Schultz Family Foundation, Cognizant and Michael and Susan Dell Foundation and piloted in collaboration with WhereWeGo, AdeptID and Brighthive. It was designed to answer user questions and direct people to relevant resources across our platform. We hoped that the experience would reduce search fatigue, increase user confidence for an already skeptical audience and accelerate the career exploration process with our highly vetted catalogs and platform. We tested it rigorously; small batches of user testing, a 500-person pilot, and a 5,000-person beta. While we had high aspirations, the results were mixed.
The chatbot worked well for simple questions, but fell short when users needed help navigating the more complex and emotional process of career change, particularly young adults and others facing systemic barriers to mobility. Decision fatigue, low confidence, lack of clarity — those are hard things to solve with just a Q&A interface.
You can read more about what we learned in a newly published report from MDRC. The report details the pilot, beta and feedback loops we used to evaluate SkillUp AI performance — all of which were strengthened through our work with rigorous partners, expert researchers and data scientists to ensure deep evaluation and insight at every stage.
From Information to Action
We needed something smarter and more advanced, in addition to the chatbot. Beyond centering AI around information retrieval, our next phase of work explores how it can act as an embedded real-time guide through our platform. With its seamless integration in our user interface, it helps users take the next step and, hopefully one day, the step after that. Ideally, we want to reduce drop-off points, eliminate guesswork and deliver personalized nudges that match where someone is in their journey.
For example, if a user is browsing training options but hasn’t enrolled, can AI surface just the right encouragement or support tool? If someone uploads a resume, can AI recommend career paths that align with their experience — particularly ones they may not have considered yet? Can AI clarify eligibility in plain language and help users market themselves with confidence?
These are both technical and human challenges. Finding a solution to both is core to SkillUp’s mission of increasing access to high-opportunity careers for STARs.
Intelligent Experiments, Grounded in Equity
As part of our ongoing work, we’re running real-time experiments inside the SkillUp platform that explore these use cases. The approach is intentionally iterative. We test small, learn fast, then build features based on actual user behavior and feedback. This might look like AI-powered recommendations in our training catalog, tailored alerts when a user hasn’t visited the site in a while, or friction-reducing features that surface career guidance in more intuitive ways. Early learnings indicate users are more engaged with prompts that remove decision paralysis and directly address their personal context — such as “Is this program a fit for me?” — compared to broader questions like “What are the pros and cons of this program?” or abstract, risk-based prompts like “What is the path to $100K?” Personalized, situational framing appears to drive greater relevance to and action by our test users.
We don’t experiment for experimentation’s sake. We do it to ensure AI actually improves user experience, especially for those who have in the past been overlooked or excluded by education and workforce tools that were not built with STARs’ needs in mind. And we’re committed to being transparent about what works and what doesn’t.
A Platform That Meets Users Where They Are
One of our core takeaways? AI in workforce development has to be designed smartly so it reflects real human needs.
The SkillUp AI shift is ambitious, aiming to create a companion experience that acts in the best interest of a user. “Many organizations are throwing AI out there without strategy. We are laser focused on intentional design that supports action,” said Jason Cheng, SkillUp’s SVP of Product and Analytics. “We don’t want to build a robot that answers questions, we want to build a guide that helps STARs move forward.”
That mindset is what sets SkillUp apart.
What’s Ahead
We aren’t done iterating yet. Our boldest step yet lies just ahead. We’ve begun to explore agentic AI that can proactively take actions on a user’s behalf like navigating a website, making recommendations or completing tasks, rather than waiting for the user to prompt it. We were among 20 organizations selected by Google — out of a pool of 3,000 applicants — to participate in the Google Generative AI Accelerator to make this step a reality.
Agentic AI could be the teammate that nudges users in the right direction, at the right time, so they don’t get stuck. We think this will be powerful for STARs and especially young people, many of whom are still forming their sense of purpose but feel overwhelmed by the choices and actions ahead.
Want to learn more? Dig deeper into the next steps for SkillUp AI in this Part 2 article!
