Gemini Skills: The Google Feature Saving Teams Hours (Almost No One Knows About It)

Every time you switch between Gmail, Docs, Slides, Sheets, and Chat, your brain pays a small tax. You lose context. You lose time. You lose focus. A 2024 study by RescueTime found knowledge workers switch between apps roughly 1,200 times a day. The actual cognitive cost is not the switch itself -- it is the re-loading of context every time you land somewhere new.

Most AI tools made this worse, not better. They added another tab, another login, another place to context-switch. The AI helped you write a paragraph, but the tax on getting to that paragraph stayed the same. You still had to find the spec doc, the email thread, the client's last reply, the spreadsheet of numbers, before the AI was useful.

Google's Cloud Next '26 announcements are an attempt to remove that tax. Not by making a smarter chatbot, but by rewiring how the apps connect underneath. The four pieces that matter for teams are Workspace Intelligence, Ask Gemini in Chat, Projects, and Skills. Long-running agents sit on top.

Workspace Intelligence is the underlying graph. Google calls it "a new work graph." It connects emails, Docs, Slides, Sheets, Chat, and Drive so that one query can pull from all of them at once. You ask "what is the status of the Acme renewal?" and Gemini stitches the answer from a Google Chat thread, two emails, a doc, and a row in a sheet. Ask Gemini in Chat is the entry point. Inside Google Chat, type what you need: a daily briefing of unread threads and action items, a meeting scheduled across five calendars, a file you can describe but cannot remember the name of. Critically, Ask Gemini in Chat connects to outside tools -- Asana, Jira, and Salesforce -- so updates from those systems flow back into Chat without tab-hopping. Google confirmed this is generally available now. Projects is shared context. Drop a project's Docs, emails, and chats into a Project, and Gemini becomes the expert on that one initiative. When a teammate joins late, they do not need three hours of catchup -- they ask Gemini and get answers grounded in the project's full history. There are two flavors: Drive Projects in Workspace, and Projects inside the Gemini Enterprise app for company-wide use. Skills is the productivity unlock. Reusable workflows you teach once and run forever. Skills are built in Workspace Studio (the no-code agent builder Google launched in December 2025) and invoked anywhere Gemini is available -- including Ask Gemini in Chat via the @skill-name pattern. Google announced Skills at Cloud Next '26 in April 2026 and described them as "rolling out in coming weeks." Long-running agents sit on top of all this. You give an agent a goal -- "research this competitor before Friday's meeting" -- and walk away. The agent plans steps, runs them, pauses to ask for approval at checkpoints, and delivers a finished output hours or days later. The Google Cloud Agent Runtime supports operations of up to seven days. Long-running agents are managed via the Inbox -- "air traffic control for your agents." Google confirmed long-running agents are generally available in the Gemini Enterprise Agent Platform.

Google's launch posts speak in capability language ("now you can"). The reality on the ground is tier-gated. Here is the honest breakdown as of May 2026.

| Feature | Status | Tier(s) needed |

|---------|--------|----------------|

| Ask Gemini in Chat (with Asana / Jira / Salesforce) | GA | Business Standard / Plus / Enterprise |

| Workspace Studio (the builder) | GA since Dec 3, 2025 | Business and Enterprise plans |

| Skills (reusable workflows in Workspace Studio) | Rolling out (April 2026) | Tier breakdown not yet officially detailed -- track the Workspace Updates blog |

| Workspace Intelligence (the work graph) | Rolling out | Tier breakdown not yet detailed |

| Long-running agents | GA (up to 7 days runtime) | Gemini Enterprise (a separate add-on) |

| Projects in Gemini Enterprise | GA | Gemini Enterprise |

| Drive Projects | Rolling out | Workspace |

| Inbox (agent control plane) | GA | Gemini Enterprise |

Two things to flag honestly. First, Business Starter is capped at five Gemini prompts per day with the standard model -- Skills, Projects, and the agent features are effectively not on Starter. Second, some of the heaviest features (long-running agents, Projects in the Gemini Enterprise app) require Gemini Enterprise, which is a separate add-on, not the same thing as your base Workspace plan.

If your domain is on Business Standard or higher, you have access to Workspace Studio today and Skills as it rolls out. If you need long-running agents that run for days inside a secure Google Cloud environment, you need Gemini Enterprise.

---

A regular Gemini prompt is a one-shot instruction. You write it. Gemini answers. The instruction disappears. Next time you need the same workflow, you reconstruct the prompt from memory or from a screenshot you saved. The team has no shared way to do it the same way -- so five people produce five slightly different outputs.

A Skill is a saved, named, parameterized version of that workflow. You build it once in Workspace Studio. You give it a name -- format-client-report, weekly-team-update, apply-brand-guidelines. You define what inputs it takes (a doc, a client name, a date range) and what it produces (a formatted report, a status email, a branded slide deck). Once published, anyone on your team can invoke it -- in Ask Gemini in Chat (@format-client-report Acme Q2), in a Doc, in a Sheet -- and get the same workflow with the same standards every time.

The difference is roughly the same as the difference between writing one-off bash commands and writing a script. The script is reusable. Five people running the same script get the same result. Five people reconstructing a bash command from memory produce five subtly different things, and the bug-hunting comes after.

Build one skill and you save the time of redoing one workflow. That is small.

Build five skills and a typical workweek starts to feel different. The repetitive multi-step work that used to fill three afternoons now compresses into running three skill invocations. The afternoon is now spent on judgment work the agent cannot do -- catching the thing that feels off, deciding what matters, calling the customer.

Build twenty skills and your team's output changes shape. The boring scaffolding of knowledge work -- weekly reports, client onboarding sequences, brand-aligned proposals, monthly retrospectives -- runs off skill invocations. Newer team members can do a senior person's first draft without senior-level context, because the skill encodes the senior-level context.

This is why Google calls Skills the productivity unlock. The unlock is not in any one skill. It is in the cumulative effect of replacing dozens of fragile, person-dependent workflows with one durable, team-shared one.

A Skill is built in Workspace Studio. Inside, a skill encodes:

• The instruction -- what the workflow does, written in plain English
• The inputs -- what the user provides each time (a client name, a date range, a source doc)
• The data sources -- which Workspace surfaces it pulls from (Drive, Gmail, Sheets, plus connected third-party tools like Asana, Jira, or Salesforce via Ask Gemini in Chat connectors)
• The output format -- the final shape (a Doc, a slide deck, a sheet, an email draft, a Chat message)
• The guardrails -- approval checkpoints, who can run it, what data is in scope

Workspace Studio handles the no-code interface for all of this. You do not write the skill in JSON or YAML. You write it the way you would explain it to a new hire, and Gemini compiles the spec.

---

This is the example Google walked through in the Cloud Next '26 keynote and Ava's video. It is the simplest concrete demonstration of why long-running agents and Skills change the shape of work.

The old way. You have a client meeting Friday. Wednesday morning you start researching the competitor. You open ten browser tabs. You read three analyst reports. You skim the competitor's blog, their pricing page, their job listings. You take notes in a doc. You build a slide. You think about what to say. Total time: most of a day, probably split across two days because you keep getting interrupted. The new way. Wednesday morning you open the Inbox in Gemini Enterprise and start the deep research agent. You give it the competitor's name and the angles you care about -- pricing strategy, recent product launches, hiring trends, customer complaints on Reddit. You walk away.

The agent runs autonomously. It pulls from public sources, summarizes findings with citations, flags conflicting reports, and pauses to ask you a clarifying question if something is ambiguous ("the competitor announced two pricing changes in March -- which one do you want the brief to focus on?"). When it is done -- typically a few hours later -- there is an executive summary in your Inbox.

Then you invoke a Skill: @build-competitor-brief-deck with the research summary as input. The Skill takes the summary, applies your company's slide template and brand style, generates a fully editable deck with speaker notes, and drops it in your Drive Project for the client account. Total time of your active attention: about an hour, mostly spent reviewing the deck and adjusting the parts where your judgment matters.

The combination is the point. Long-running agent does the grinding. Skill does the formatting. Project gives both of them the shared context they need. Ask Gemini in Chat is where you kick the whole thing off and where the daily briefing surfaces the result.

This one is mundane. That is why it matters. Mundane workflows are the ones that consume the most cumulative time.

The old way. Every Friday afternoon, the team lead writes a weekly update. They open Google Chat to find what shipped. They check Jira for closed tickets. They look at Asana for completed tasks. They glance at the analytics dashboard. They open last week's update doc to match the format. They write the new one. They send it. Total time: 45 minutes to an hour, every week, forever. The new way. The team lead builds a Skill once: @weekly-team-update. The Skill is configured to:

1. Pull closed Jira tickets from the past seven days
2. Pull completed Asana tasks from the past seven days
3. Pull recent updates from the team's Google Chat space
4. Pull last week's update Doc as the format reference
5. Generate this week's update in the same format, grouped by workstream
6. Drop the draft in the team's Drive Project

Friday afternoon the team lead types @weekly-team-update in Ask Gemini in Chat. Forty-five minutes of work compresses into about four minutes of editing the draft and sending it. Multiply that by 50 weeks a year and the skill returns roughly 35 hours of senior time annually -- to one person. If three team leads share the same Skill, that is over 100 hours back to the team per year, from one Skill.

The Asana / Jira / Salesforce connectors that ship with Ask Gemini in Chat are what make this Skill possible. Without them, you would have to manually export status from each tool every Friday. With them, the Skill pulls live data on every invocation.

---

Pick a project you are actively working on. A client engagement, a product launch, a hiring round -- something with at least three Docs, a few email threads, and a Chat space. Create a Drive Project (or Project in the Gemini Enterprise app, if you have it). Drop the relevant files in. Connect the email threads. Add the Chat history.

Now ask Gemini questions about the project in plain English. "What did the client push back on most?" "Who owns the next deliverable?" "What is the current open question?" The point is to feel out how it remembers context across surfaces. Most teams underestimate how much context they have scattered until they see it consolidated.

The deep research agent is the easiest place to start. Pick a research task you have been putting off -- a competitor analysis, a market sizing, a hiring landscape review. Give the agent the task and the angles. Walk away. Come back later and look at what it built.

Two things will happen. First, you will be surprised at the quality of the executive summary and the citations. Second, you will find at least one place where the agent missed nuance only a human would catch. That second part is the point. The new shape of work is humans doing judgment, not humans doing grinding.

For one week, start every workday by typing a daily briefing prompt into Ask Gemini in Chat. "What needs my attention today?" Let it scan your inbox, calendar, and connected tools (Asana, Jira, Salesforce). Read the briefing. Decide what to actually work on.

This habit alone changes how mornings feel. Most knowledge workers spend the first 30 minutes of every day in a triage scramble -- chat, email, calendar, three tabs of notifications. The daily briefing replaces that with a one-page summary and an explicit prioritization. Same information, less cognitive cost.

Pick a workflow you do at least weekly. Examples: the weekly team update, a client onboarding sequence, a monthly invoice reconciliation, a recurring brand-aligned proposal. Open Workspace Studio. Build the Skill. Test it. Share it with the team.

Use it for two weeks before building another one. The temptation is to overhaul everything in a sprint. Resist. Each Skill is a piece of organizational memory -- it needs to be tested, debugged, and refined against real use before you trust it. Build one, get good at it, then add the next.

After three months at this pace, you will have a small library of well-tested Skills covering the workflows you actually do. After six months, the team's week looks materially different from the way it looked before April 2026.

---

A few things Google has not pinned down publicly as of May 2026:

• Skills tier breakdown. Google has confirmed Skills are rolling out, but has not fully detailed which Workspace plans get them and at which limits. Track the Workspace Updates blog before committing a team to Skills as a critical workflow.
• Cross-domain Skill sharing. Whether Skills can be shared across organizations (for example, by a consulting firm to a client) or only within a single Workspace domain.
• Skill governance. What controls exist for who can create vs. who can run a Skill, especially when it touches Salesforce or Jira data.

A few claims around the launch deserve a skeptic's eye:

• "Build five skills and your week looks completely different." True if you pick the right five workflows. Less true if you build five novelty skills and skip the boring ones that consume real time. The framing is correct in spirit but assumes good selection.
• "AI agents are great at repetitive multi-step work, bad at judgment." This is the line most launch coverage repeats. It is correct as a current snapshot, but the boundary moves. The agents that ship in May 2026 are noticeably better at judgment than the agents that shipped in November 2025. Reassess every quarter.
• "Your data stays your data." Google's policy is clear that Workspace customer data is not used to train models outside of the customer account, with no human review. Verify the latest version of the policy in your admin console before putting client data in a Project.

Two things to watch over the next two quarters:

1. Skill libraries. Whether teams start sharing Skills the way developers share npm packages, with a public catalog. Google has hinted at an Agent Gallery in the Gemini Enterprise app -- whether that extends to Skills built in Workspace Studio is the question.
2. Skill cost. Skills consume Gemini quota on every invocation. As teams adopt them and run them dozens of times a day, the per-tier prompt limits become the bottleneck. Watch for tier changes from Google through 2026.

The bigger picture is that Workspace is no longer a tool you operate. It is a platform that operates on your direction. The teams that figure out how to direct it well -- one Skill, one Project, one daily briefing at a time -- will compound their output through the rest of the year. The teams that keep pasting one-shot prompts into a sidebar will keep doing the same week of work next week.

---

_Source: Google's New Gemini Update Is INSANE -- Ava on YouTube. Verified against Google's official Cloud Next '26 announcements: 10 more announcements for Workspace at Next 2026, The new Gemini Enterprise: one platform for agent development, New Workspace Intelligence, and Now available: Workspace Studio (Dec 3, 2025)._