Autovance

Powered by ABR technology

Book a call

AI Skills and Always-On Agents: How to Turn Your Team’s Know-How Into Repeatable Work

AI News ·

AI Skills and Always-On Agents: How to Turn Your Team’s Know-How Into Repeatable Work

In the first half of November, two announcements pointed in the same direction. On 4 November, Perplexity launched Teammates (opens in a new tab), AI agents designed to keep working on a job over days and weeks rather than answer one question. On 10 November, Google confirmed that skills in Google Workspace (opens in a new tab) are rolling out, which let Gemini follow a team’s saved rules, templates and reference files inside Gmail, Docs, Slides, Drive and Chat.

Neither is a big-bang product for small businesses. But together they show where AI at work is going: away from clever one-off prompts and towards written-down ways of working that the AI repeats every time.

For an owner, that shift matters more than any single feature. The businesses that get the most from these tools will be the ones that know, and have written down, how they do things.

What changed

Google: reusable instructions inside everyday apps. According to Google’s Workspace Updates blog, a skill is a reusable set of instructions that guides Gemini to follow your team’s rules, templates and reference files. You can build one in Workspace Studio, draft it together in Google Docs with a Skill Builder, or ask Gemini to create one while you work. To use it, you mention it with @ in the Gemini panel, or Gemini applies it when it judges it relevant. Skills can also start Workspace Studio flows, either by hand or on a schedule.

The details matter for planning:

  • Eligible editions include Business Starter, Standard and Plus, as well as the Enterprise editions.
  • Rollout started on 9 November for Rapid Release domains and starts on 30 November for Scheduled Release domains. Google says each can take up to 15 days to become visible.
  • Skills are on by default for eligible editions. Admins can switch them off for the whole domain, a department or a group.
  • They don’t work in Google Sheets or Google Meet yet, and prompts and instructions must be in English.

Perplexity: agents that own an ongoing job. Perplexity’s announcement describes Teammates as personal agents for long-running work. They keep the project’s goals, task history and files between runs, watch for changes, act on a schedule you set, such as a weekly pipeline review, and learn from your corrections. In Perplexity’s words, “Personal agents need to be purpose-built for work.” Teammates ask for approval before completing work. For now they are available on the web for Perplexity Max subscribers only.

What it means for a small business

Most small teams already use AI, but usually one person at a time, each with their own prompts. The results vary from person to person, and the good prompts live in someone’s chat history.

Skills fix part of that. When your best way of writing a quote follow-up, a client update or a proposal is saved as a skill, everyone gets the same starting point. A new hire gets your house style on day one. And when you improve the instructions, you improve them for the whole team.

Agents like Teammates go a step further: they don’t wait to be asked. That is powerful for routine checks, such as chasing overdue paperwork or flagging a drop in enquiries, but it raises the stakes. An agent that acts on its own needs clear rules, limited access and someone who reviews its work.

The common thread is simple. AI is only as consistent as the instructions you give it. If your processes live in people’s heads, these tools can’t help much. If they are written down, the tools can repeat them reliably.

Practical steps for this month

  • List your repeat tasks. Spend 30 minutes with your team noting the written work you do every week: replies to common enquiries, follow-ups, status updates, meeting summaries.
  • Pick three and write them down. For each, write one page: when it happens, what good looks like, which template or files to use, and what must never be said. This is the raw material for a skill in Google, Microsoft or any other tool.
  • Check your Workspace settings. If you use Google Workspace, ask your admin whether your domain is on Rapid or Scheduled Release, and decide who should have skills switched on first.
  • Test before you share. Run each skill on five real examples and compare the output with what your best person would write. Adjust the instructions, not just the output.
  • Name an owner. Each skill or agent needs one person who keeps it up to date and decides when it is retired.
  • Keep a person in the loop. Anything that reaches a client, a supplier or your accounts should be approved by a human until you trust the results.

Pitfalls to watch

Copies that drift apart. Google says that when you share a skill, the recipient gets an independent copy, and later edits to the original don’t reach copies already saved. If five people copy your quote skill and you then change your prices, you have five outdated versions. Keep one master document and tell people when it changes.

Files the AI can’t see. A skill that points to a price list or template in Drive only works if the person using it has access to that file. Check permissions when you share.

Automating a broken process. If your current follow-up routine is messy, an AI will repeat the mess faster. Fix the steps first, then automate them.

Too much access, too soon. An agent that can read your inbox and send messages can do damage as well as good. Start with read-only tasks and short checklists, and widen access slowly.

Old setups. Google says skills don’t sync with Gems in the Gemini app, and remaining Gems will be migrated to draft skills in 2027. If your team built Gems, note where they are used so nothing breaks quietly.

The takeaway

The new wave of AI tools rewards businesses that are organised. Saved skills and always-on agents will repeat whatever you teach them, so the most valuable work this month is not choosing a tool but writing down how your team does its best work, with clear owners and approval steps.

If you’d like help deciding which tasks to hand to AI and how to set them up safely, see how we approach AI and workflow automation. And if those tasks depend on data spread across several apps, we can also connect your business systems so the AI works from one source of truth.

Sources: Perplexity blog (4 November 2026), Google Workspace Updates blog (10 November 2026).

Frequently asked questions

What is an AI skill?

It is a saved set of instructions that tells an AI assistant how your business does a particular job: the rules to follow, the template to use and the reference files to check. Instead of typing the same long prompt every time, you call the skill by name, and everyone on the team gets the same approach.

Do we need Google Workspace Enterprise to use Gemini skills?

No. Google lists the Business Starter, Standard and Plus editions as eligible, alongside the Enterprise editions and two add-ons. The feature is rolling out gradually from 9 November for Rapid Release domains and from 30 November for Scheduled Release domains, so it may take a few weeks to appear in your account.

Is it safe to let an AI agent work on its own?

Only within limits you set. Perplexity says its Teammates ask for approval before finishing work and run inside the permissions you give them. Whatever tool you use, start with tasks where a mistake is cheap, keep a person approving anything that goes to a client, and give the agent access only to the data it needs.

Where should a small business start?

Pick two or three tasks your team repeats every week, such as answering a common enquiry or writing a status update. Write down how your best person does each one on a single page. That page is the raw material for an AI skill in any tool, and it is useful for training new staff even if you never automate it.

Want this working in your business?

Tell us what you are trying to fix and we will show you what it would take.

Book a call