Google Replaces Gemini Gems With Skills for AI Workflows

Google is retiring Gemini Gems in favor of a new feature called Skills, a change that points to a broader shift in how users configure AI assistants. Gems let people create personalized assistants with custom instructions. Skills are designed to provide a similar layer of customization, but with a more reusable, workflow-oriented shape.

From personalized assistants to reusable instructions

According to The Verge, Google plans to replace Gems with Skills starting November 17. The practical difference is less about a new model and more about how behavior is packaged. Instead of treating every customized assistant as a separate destination, Skills can act as repeatable instruction sets that users apply to Gemini tasks.

That distinction matters for teams building automations. A custom instruction set is easier to version, document, test, and reuse than a one-off assistant configuration. It also maps more naturally to the way modern agent systems are assembled: a model, a set of tools, a knowledge layer, and a bounded operating procedure.

Why builders should pay attention

The move also reflects growing pressure to make AI systems more composable. The most useful business agents rarely depend on a single prompt. They combine specialized capabilities such as retrieving information, transforming data, calling APIs, and handing work to another step. Reusable Skills could give less technical users a simpler entry point into that pattern, while developers continue to manage more explicit orchestration underneath.

For automation teams, the important question will be whether Skills can be shared across users, environments, and connected services. If they remain tied to one account, they will be convenient personalization features. If they become portable and governable, they could become building blocks for larger workflows.

Key takeaway: Google’s shift from Gems to Skills is a small product change with a larger implication: AI customization is moving toward reusable components, which is the same direction agent and automation builders are already taking.