SHRM26 Debrief · AI, Data & Tech · #3582
Curate. Discover. Converse. How AI and Human Expertise Are Redefining Learning Discovery for HR
with Yinglu Chen
"The wheel exists. The wheel runs."
What it was about
Amazon's PXT Academy built an AI-powered, people-centered learning ecosystem for 8,000 HR professionals across 30 countries by shifting from a one-size-fits-all 'learning journey' to a personalized, self-directed 'learner journey.' It uses AI to curate, validate, and generate knowledge at scale while keeping humans (experts, librarians, managers, mentors) in the loop at every stage.
By the numbers
90% reduction in content rebuild time
Efficiency gain from the AI-powered curation/build process
Four hours to 22.5 minutes
Time reduction for a LARA-related content task
At least three weeks before start date
Recommended manager touchpoint timing for new-hire onboarding prep
Key notes
- Start with curation, not creation — map existing content against a capability framework before building anything new, since most organizations already have usable assets buried in an unwieldy LMS.
- Treat AI-generated content as a draft that always requires human expert review before it reaches learners, to build a validated 'golden dataset.'
- Separate development tools from performance management systems so learners don't perceive AI-driven learning recommendations as punitive or surveillance.
The contrarian takeDeliberately deciding NOT to integrate the AI-driven learning/development system with performance management, even though doing so could make ROI more visible — because tying development tools to performance review causes learners to perceive them as punitive rather than supportive.
Take this back Monday
Do this for your team
Map your team's existing training materials to skills gaps before creating anything new — curate what you already have first.
Say this in your next leadership meeting
AI plus human expertise equals ROI — we should use AI to curate and personalize learning, not just generate more content.
Watch out for
- Treating a growing content library as the goal ('more content') when the real issue is a connection problem, not a content problem — bigger libraries increase learner disengagement.
- Assuming all learners at the same level need the same 'peanut butter spread' curriculum instead of a personalized path based on individual capability gaps.
- Recreating content that already exists (e.g., translated/localized assets) instead of curating and mapping what's already working locally.
If this landed, see these
↳ Go deeperFuture-Ready HR: Leveraging AI to Drive Learning, Growth, and Talent MobilityGives individuals a personal framework for turning AI-curated learning into actual career mobility.⇄ The counterpointState of the Workforce 2026: What to Keep, What to Change, and What AI Is Actually Doing to Your EmployeesA massive workforce dataset says core engagement drivers barely move, however personalized the tech gets.✦ The unexpected oneThe Benefit That Quietly Shapes How Work Gets DoneSame lesson, different budget line: a commoditized-seeming line item hiding outsized, overlooked value.