The next platform won't fix what platforms can't fix.
A new wave of evidence is making a buying case for AI-personalised learning. A longer line of implementation research keeps finding that the platform itself is less than a third of whether it lands. What CHROs and CTOs should ask before the next purchase.
A wave of peer-reviewed research is making a buying case for AI-personalised learning. Studies from 2024 and 2025, including a 2025 survey of 268 university instructors, report measurable improvements in learner performance and engagement when machine learning systems adapt content, pacing, and difficulty to the individual. The pattern holds across different cultures and platform generations. Most CHROs will see a vendor deck citing some version of it before the quarter is out.
A longer line of evidence sits underneath that wave, and it has been saying something different for a decade. A systematic review of LMS use in higher education, plus studies applying the DeLone and McLean information systems success model and the LPOMR (leadership, planning, organisation, management, resources) framework, all land in the same place: less than a third of whether a learning platform works is the platform itself. The rest is sponsor cover, who owns the boundary between the tool and the work, change management that survives a reorganisation, and what gets rewarded in practice when no one is reporting on it. That finding has held across institutions, regions, and three generations of platform.
The two streams of research sound like the same conversation, but they are doing different jobs. The personalisation studies report what happens when an individual learner uses a tool that is already configured, sponsored, and built into a workflow that matters. The implementation studies report whether any of that setup ever exists in the first place. Most buying decisions take the first answer to a question only the second answer can settle.
The implementation literature names the same elements every time. You need a sponsor with a stake in the outcome, not just sign-off on the budget. You need named owners for the boundary between the platform and the workflow it is meant to change, with clear escalation for when the two meet and one of them gives. Change management has to be treated as design, not as communications. Governance has to be written down: what knowledge the model can reach, who can revoke that access, and what evidence would justify revoking it. And you need capability signals the executive team will accept, which means signals that do not come from the vendor's own dashboard.
A few asks decide whether you repeat the pattern or break it. First, a diagnosis in writing of why the last platform under-delivered: a specific account of which conditions were missing and which leaders did not move when the system asked them to, rather than a story about the vendor. Second, a small set of non-platform capability signals you can stand behind, of the kind a workforce data strategy tends to land on. Third, a sponsor pair, the CHRO and the CTO or CIO, committed in writing to the changes the platform will surface, not just to the platform itself.
This is the work a capability diagnostic does in six weeks. It spends less time on the tool and more on the ecosystem the tool is dropped into: what the organisation already teaches by default through incentives, calendar pressure, and who gets believed in a meeting; where the proxies are lying; where the previous procurement quietly lost momentum and why nobody quite said so at the time. The output is in plain language, because the next conversation it has to survive is a leadership forum with L&D out of the room.
For senior L&D leaders, the new personalisation evidence makes the strategic mandate clearer, not weaker. AI-personalised learning is a real tool with real lift, in the right ecosystem. The market will reward vendors who can show strong individual-outcome data on adaptive pathways. It will not reward L&D functions willing to argue that no platform produces capability in conditions it was never set up for. That argument is where the function earns its place in the executive room. If that is your remit and you want a direct conversation about what those conditions look like in your organisation, submit a brief and we will tell you in writing whether we are the right fit.
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 diagnostics, and the workforce data that survives a board meeting.