In many cases, new joiners are expected to contribute quickly, often in roles where core skills are built through doing rather than instruction. Most onboarding programs reflect this. They rely on cases, simulations, and team-based exercises to mirror the work itself. Yet one constraint persists: Feedback, the mechanism that helps people improve between attempts, is limited.
Facilitators may not be able to observe every individual closely. Peer input varies, especially when everyone is new. Feedback often comes after the moment when it could shape the next attempt. People improve, though the path is not always as direct as it could be.
Recent advances in AI have opened up a different possibility for us. They made it feasible to provide immediate, individualized feedback at scale during practice while also giving a clearer view of how learners were progressing. This created an opportunity to strengthen deliberate practice—structured, personalized, feedback-driven skill building—without changing the overall structure of onboarding.
In an earlier People in Progress post on moving from orientation to impact, we explored how onboarding can better prepare new joiners for real contribution. Building on that work, we looked more closely at how feedback, standards, and facilitation could evolve when these capabilities were introduced.
Bringing feedback into the moment: Feedback became part of the task itself
Learners continued to work through the same types of problems. As they did, they could test their thinking, receive input, and adjust while still in the exercise. In practice, this meant a learner could submit a draft response or recommendation and receive immediate, structured feedback aligned to defined skill criteria. In a problem-structuring exercise, for example, feedback could point to gaps in logic, missing drivers, or unclear framing, allowing the learner to revise before moving on.
As we introduced this, the rhythm of learning began to shift. Shorter cycles of attempt and refinement replaced longer gaps between effort and feedback. Progress became easier to see across participants, and learners could connect feedback directly to the choices they had just made. The timing of feedback began to carry more weight in how learning unfolded, reinforcing a more deliberate way of building skills.
Making standards explicit: More frequent feedback raised a practical question about what that feedback should be anchored in.
In many programs, expectations of strong performance are understood but not always spelled out. As we worked through this, we found it useful to describe core skills in more concrete terms. What does effective problem structuring look like in practice? What makes a recommendation clear and well supported?
Simple rubrics helped apply these expectations across different exercises. They created a shared reference point for learners, facilitators, and the system—one that carried across interventions, not just within a single session. Over time, feedback became more specific, more consistent, and easier to build on from one learning moment to the next.
We also saw that how feedback was integrated mattered. When it appeared within the task, learners engaged with it as part of their work. When it sat outside the workflow, it was easier to overlook.
What changed in the room: Some of the shifts were most visible in how sessions unfolded.
Learners spent less time waiting to understand how they performed. They could act on feedback while the task was still fresh, making it easier to improve in the next attempt.
Facilitators had a broader view of how the group was progressing. Aggregated signals made it easier to spot where learners were converging or struggling, which enabled real-time adjustments to the session. In some cases, examples of strong responses helped anchor the discussion and make expectations more tangible.
This also changed how time was used. Less time was spent on broad, retrospective feedback and more on targeted coaching and deeper discussion.
For the learning team, the experience created a more consistent view of skill development across cohorts. This made it easier to refine the program based on observed patterns rather than isolated examples.
What it took to make it work: A few practical considerations shaped how this came together.
Building trust proved important. Many learners are already familiar with AI tools but are often cautious about automated feedback, particularly how well it reflects expert judgment and role expectations. Feedback landed more effectively when it was clearly grounded in expert standards and tied to what good performance looks like in the role.
Sequencing also played a role. We asked learners to complete the task on their own first, then review feedback, and then refine their work. This helped maintain focus on building core skills while still allowing learners to benefit from targeted input. In some cases, learners then used additional tools to further improve their output, building an understanding of how different forms of input can shape quality.
Simplicity mattered, particularly given how onboarding is delivered. Facilitators may often rotate in and out, with limited time to prepare. Clear guidance and a small number of consistent practices made adoption more manageable.
Looking ahead.
This work started as an effort to improve a specific part of onboarding. It has since influenced how we think about learning more broadly.
Stronger feedback loops have made it easier to see how skills develop in real work and where adjustments are needed. They also enable a more continuous way to refine how learning shows up in the flow of work, based on what people are actually doing day to day.
We are building on this by using these insights to spot where support is needed most—including differences across regions—and to tailor interventions beyond the first weeks, focusing on the moments that matter for performance.
One question has stayed with us: Where could more immediate and consistent feedback change how quickly people build capability?