Manager conversations
Guidance helped managers support development without needing to become AI experts themselves.
01
AI CAPABILITY · CASE STUDY
AI NAVIGATOR
Moving employees from “How do I use AI?” to “How can I redesign my work?”
OVERVIEW
As generative AI adoption accelerated, employees needed more than tool training. They needed a way to understand their own capabilities, make sense of where AI could actually help them, and build the judgment to apply it responsibly in real work.
I designed AI Navigator as an enterprise capability-development experience built around an agentic assessment and coaching tool, then created the surrounding strategy needed to turn personalized recommendations into continued learning, experimentation, and behavior change.
THE CHALLENGE
Build durable capability, not feature training.
AI learning can become obsolete almost as quickly as it is created. A course explaining features or prompting techniques might help someone today, but it does not necessarily help them decide when AI is useful, recognize the quality of an output, adapt their approach when something fails, or rethink the way work gets done.
How do we help employees develop durable AI capability rather than simply learn how to use an AI tool?
WHAT I FOUND
One learning path would not fit everyone.
Employees were starting from very different places. Some needed foundational understanding. Others were already experimenting and needed help evaluating results, designing better workflows, or recognizing more sophisticated use cases.
The experience needed to meet people where they were, recognize capability across multiple dimensions, connect development to real work, provide useful next steps immediately, and support continued growth rather than end when a course was complete.
01ASSESS
→
02PERSONALIZE
→
03APPLY
→
04REASSESS
WHAT I DESIGNED
An agentic assessment and coaching experience.
I designed an assessment spanning six dimensions of workplace AI capability. Instead of returning a score and a list of courses, the AI coach interpreted the employee’s responses, explained its reasoning, identified development opportunities, and recommended activities based on the individual’s current capabilities.
The goal was not to tell someone where they ranked. It was to help them make an immediate developmental shift.
WHY THIS APPROACH
I was more interested in movement than labels.
The value of knowing someone’s current capability is limited unless that information changes what happens next. So the assessment became the starting point for a personalized development conversation rather than the final product.
That also changed the role of AI. It was not simply being used to generate learning content faster. It was part of the learning architecture itself, interpreting information, adapting recommendations, and creating a more responsive experience for the learner.
THE SYSTEM AROUND IT
The coach was the core. The surrounding system made development stick.
Peer learning
Employees could learn from how colleagues were experimenting with AI in different roles and workflows.
Workflow experimentation
Development activities encouraged people to apply AI to actual work rather than only hypothetical exercises.
Curated resources
Recommendations connected people to relevant learning and support instead of sending everyone through the same content.
Reinforcement
The experience encouraged continued practice rather than a single interaction.
Reassessment
Employees could revisit their capabilities over time and identify the next useful area for development.
AGENTIC CORE → PERSONALIZED DEVELOPMENT → SURROUND STRATEGY → APPLICATION → REASSESSMENT
RESPONSIBLE USE
Capability with AI is also about judgment.
Employees needed to evaluate outputs, recognize limitations, protect sensitive information, and understand when human review was essential. I partnered with Legal and Privacy colleagues on responsible-use guidance and connected AI Navigator to a broader AI learning ecosystem, including AI Literacy: Understand, Evaluate, Create, which I designed and built in Articulate Rise.
The intent was to make responsible judgment part of capability development rather than a compliance message bolted onto it afterward.
ITERATION
The first version was a starting point, not the finish line.
After launch, I used participant feedback and performance signals to refine the experience. I evolved Minerva from a primarily diagnostic interaction toward a more repeatable development loop: listen to role context, current habits, and confidence; diagnose AI work patterns and judgment; recommend next-best practice tied to real work; and reinforce development through manager dialogue, peer support, workflow experiments, and reassessment.
That shift kept the experience focused on continued capability growth rather than a one-time assessment result.
LATER PERFORMANCE SIGNALS
later assessment-accuracy signal, equivalent to 4.05 out of 5
later perceived value after the experience was refined
percentage-point increase in perceived value, from 77% to 86%
Earlier launch feedback also showed strong trust in the coaching experience: 82% said the coach’s reasoning made complete sense, and 86% said they were likely to try the recommended development activities. Follow-up documented employees applying recommendations in their work, providing initial evidence of behavior change beyond the learning environment.
The goal was never completion. The goal was application.
WHAT I LEARNED
The most useful AI learning does not necessarily look like training.
It can look like assessment, coaching, experimentation, reflection, performance support, peer learning, or a conversation with a manager.
Start with the behavior or performance you want to change, then decide what kind of experience will actually help someone get there.
Technology expands the options. It does not replace the judgment required to choose among them.
WHAT I’D BUILD NEXT
Connect individual development even more tightly to workflow change.
I would continue refining reassessment, look for patterns across capability data, and use those patterns to identify where teams might benefit from shared practice, manager support, process redesign, or new performance resources.
The long-term opportunity is bigger than teaching people to use AI well. It is helping people recognize when AI changes what good work can look like.