AI can help build an LMS. The hard part starts after the demo

AI coding tools are making it easier to build a learning management system in-house, but a quick prototype can leave an organisation with years of maintenance and security work. In a Forbes essay, learning technology executive David James argues that extending an existing platform may be wiser than trying to replace it.

Watch Desk analysis

What happened

James, chief learning officer at 360Learning, says AI tools can produce features such as dashboards, enrolment flows and branded interfaces over a weekend. But running a platform also means handling bugs, security patches, compliance updates, accessibility and integrations as other systems change.

He recommends using AI at the edges of an established learning platform rather than replacing it outright. One example he offers is connecting AI tools to existing systems through Model Context Protocol (MCP), so a tool could use learning-completion data to inform a coaching prompt. Read James’s essay in Forbes.

James cites a 2025 MIT Media Lab NANDA study of roughly 300 enterprise AI deployments, saying vendor-built tools succeeded at about three times the rate of internally built ones and that 95% of enterprise generative AI pilots failed to produce measurable financial returns. Those are figures as presented in his essay, not a fresh assessment of the study here. He also cites a Gartner forecast that more than 40% of agentic AI projects could be cancelled by the end of 2027.

Why it matters

AI can lower the effort needed to make a first version. It does not automatically provide the people or processes needed to keep that version secure, compliant and useful. For an organisation weighing build against buy, the ongoing owner and maintenance budget belong in the calculation alongside speed and control.

James’s recommendation is not to avoid experimentation, but to put it where it can add useful capabilities without making a small team responsible for an entire platform. That is a less glamorous story than building an LMS in a weekend, but weekends are not a maintenance strategy.

Our read

A working demo can settle whether a feature is possible. It cannot settle whether the organisation should own it for years. Before building, name who will maintain the system and account for that work; if the answer is “someone on the team”, the project has acquired an invisible price tag.

Extending a proven platform may be the more practical route, provided the integration is secure and genuinely improves the learning experience. The useful test is not whether AI can make the build look easy, but whether the finished tool earns its keep after launch.

What to watch

• Whether organisations budget for ongoing maintenance, security and compliance when they build with AI.
• How AI integrations with existing learning platforms handle access to learner data.
• Whether the cited study’s findings hold across different kinds of internal software projects.

Discussion spark: When AI makes an internal tool cheap to build, should organisations still favour a vendor platform unless they can fund a permanent team to maintain their own?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.