Personalised learning, and the harder question your board is asking.
New peer-reviewed studies show AI-personalised learning lifts performance and engagement at the individual level. That is the easier half of the answer a board now wants. The harder half is yours.
A wave of studies is making the case for AI-personalised learning. A 2025 survey of 268 university instructors across Pakistan found strong correlations between AI-based personalised learning systems and improved student performance, engagement, and participation. Earlier work shows the same shape: machine learning analysing performance in real time, adjusting difficulty, pacing, and presentation to the individual learner. The pattern has held across cultural contexts and education levels. As enterprise vendors point to it, the evidence is now strong enough to put on a slide.
For a CHRO, the slide is the easy part. The harder part is what the evidence does not say.
Peer-reviewed studies of personalised learning report outcomes that platforms can record: course-level completion, time-on-task, post-test scores, self-reported engagement. Those are useful proxies for whether a tool works for a learner. They are not the answer to the question your board has actually been asking, which is whether your people can do the work the strategy now requires, under pressure, together. That question lives one level up from any platform, in how decisions are made, how disagreement is surfaced, and how a team coordinates when the script runs out.
It is the local optimisation trap by another name. Improving the learner-platform interaction in isolation is genuinely useful. It can also leave the conditions for shared judgement untouched, or weaken them, because every additional minute on a personalised path is a minute spent away from peer review, real handoffs, and shared cases. The studies record the first effect cleanly. They report nothing on the second.
A practical reading of the new evidence is to accept it where it lands, and to refuse the implied conclusion. Accept that AI tutors and adaptive pathways do useful work for individual capability. Refuse the next sentence on the vendor deck, which is that organisational capability will follow. Capability uplift is not the sum of well-tutored individuals. It is the ability of a team to act consistently under real conditions, and that emerges from the system around the work, not from the platform inside the work.
What does this mean for the slide your CHRO has to take to the board next quarter? Three things. First, distinguish individual evidence from organisational evidence and report both. Use the platform numbers honestly; show what they cover and what they do not. Second, measure a small set of capability signals that exist outside the platform: time-in-queue between named owners, exception-handling latency, rework rates on the work the team owns, or how quickly a new starter is operating to standard. These are the signals our workforce data strategy work tends to land on, because they survive contact with a board. Third, put the personalised tool inside an ecosystem that already values shared sensemaking. Where the team has time for case discussions, peer review of real artefacts, and standards that are written down and revisited, personalisation lifts the curve for everyone. Where those conditions are absent, personalisation produces the same fragmentation a content library produced, only faster.
On the technical side, the same logic applies. Where adaptive learning vendors connect to your operational tools, govern that connection in writing. Decide which knowledge sources the model can reach, who owns the boundary, and what evidence you would need to switch the connection off. That through-line shows up in our team AI capability work and in any capability strategy and build: connection without governance is not an integration, it is a hope.
For senior L&D leaders, this is also a positioning argument. The market will reward vendors who can produce strong individual-outcome data. It will not, on its own, reward L&D functions that can frame the harder question for the executive team and answer it. That work is the strategic seat you have been told you should hold. The new evidence does not threaten it. It sharpens it. The personalised tool is now a measured tool. The judgement of where it fits in the ecosystem, and how to measure what it does not measure, is yours.
Midnight Labs designs the social, technical, and environmental conditions that let organisations learn through work, not separately from it. We work with CHROs, CTOs, and L&D leaders on ecosystem design, capability strategy, and the workforce data that survives a board meeting.