If you use AI regularly for work, you have probably ended up in one of those tedious prompting loops where you spend precious, and frankly unproductive, minutes explaining the same thing again. Paste in the writing rules. Explain what belongs in the board update. Describe how the slide deck should be structured. Remind Gemini which sources you trust, what tone you want, and what it should avoid. Then come back a few days later and do it all over again.
In the earliest days of workplace AI, many of us built our own workarounds: prompt libraries, instruction blocks sitting in Notes, or old conversations bookmarked because somewhere inside them was the prompt that finally worked. For regular AI users, repetitive prompting has gradually become another form of admin. Google is now trying to remove some of it. The company has also been pushing Gemini further into everyday knowledge work, including through Gemini 3.7 Flash, while Skills addresses a different problem: remembering how users want recurring work done.
On September 30, Google started rolling out Skills in Gemini. Skills let users save instructions for tasks they expect to do repeatedly, then bring those instructions into another conversation instead of rebuilding the prompt from scratch. You can create a Skill manually, turn an existing Gemini conversation into one, or include reference material such as text files, PDFs and images. Gemini can also recognize when a Skill is relevant and apply it automatically, while several Skills can be used together in the same workflow.
That last part is particularly useful. Imagine you have one Skill that captures how you like to write and another containing your company’s brand guidelines. Gemini can draw on both when you ask it to draft a piece of communication. It sounds like a modest product update, but for people who use AI throughout the working day, it addresses one of the more irritating frictions: continually teaching the model something you have already taught it. That problem goes beyond Gemini. Palette, for example, is building around reusable AI skills and shared context, allowing organizational knowledge and working preferences to persist across different AI agents and sessions.
Most of us are not opening Gemini every morning because we want to invent an entirely new way of working. Quite often, we have already figured out what works. We just want the model to remember it. Take a founder preparing for an investor meeting. There is probably a rhythm to those meetings by now: the numbers that need to be ready, the assumptions worth challenging, the questions nobody particularly wants to answer but knows are coming, and the talking points that need tightening before walking into the room. Once that approach works, there is little value in explaining the entire process to Gemini again before the next meeting.
The same is true for communications teams and managers. After enough launches, a communications team knows how the company should sound and which phrases make everyone in the room wince. Maybe those guidelines live in a document, maybe in somebody’s Notes app, or maybe they get pasted into Gemini every time someone needs help with a launch post. Managers have their own version of the problem. Weekly reports often follow a familiar structure, as do monthly presentations and research briefs. The same pressure is appearing across AI workflow automation, where companies are trying to remove repetitive coordination and routine operational work from employees’ days. 50skills, for example, is applying that idea to recurring HR processes such as onboarding, approvals and employee administration.
Google’s own guidance for Skills focuses on exactly this kind of work. It recommends creating them when you find yourself repeating prompts, repeatedly correcting formats, or performing the same multi-step process. A Skill can hold formatting rules, templates, preferences, and common mistakes you want Gemini to avoid.
I can already think of a few I would want: one for the monthly presentation I make, one for a specific kind of report, and one for any piece of copy that has to follow five rules I have explained enough times already. That is probably the easiest way to think about Skills if you are already using Gemini at work: find the places where repeating yourself annoys you and start there.
There is also a bigger product change sitting behind the announcement. Google plans to transition Gems, the customized versions of Gemini it previously offered for repeatable tasks, into Skills. The transition starts with personal accounts in November 2026, followed by Workspace business, enterprise and nonprofit accounts in March 2027, and education accounts in June 2027. Existing Gems will be recreated as Skills. Google says Skills are already rolling out to people over 18 using personal Google Accounts, with work and school accounts coming later. Some Gems capabilities, including sharing, adding Google Drive files and using notebooks from Gemini Notebook, are also expected to arrive in Skills.
The shift from Gems to Skills says something about how Google sees everyday AI use developing. A separate custom assistant can be useful, but for a lot of day-to-day work, the more practical tool may simply be a set of reusable instructions that follows you into the conversation you are already having.
If you use AI regularly, how much of what you type is genuinely new, and how much of it is an instruction you have already given before? The cost is not only the minute or two spent finding an old prompt and pasting it back in. It is the interruption, the context switch, and the small amount of attention spent setting up work you already know how to do. AI is supposed to give some of that time back. The easiest place to start may be with the instructions you should not have to think about anymore.
