The Rise of Always-On AI Assistants, and What It Means for You
The next AI skill looks a lot like managing a team.
Picture opening your phone on Monday morning. The chat list shows: "Inbox sorted. 3 need you." "Drafted 8 follow-ups to last week's leads. Need your yes." "14 receipts in. Still missing 2."
You set these assistants up last week, and they kept working overnight while you slept. An always-on agent is an AI assistant with its own computer, tools and memory that keeps working after you leave and messages you when it needs a decision.
We are actually getting into this new reality today. Three of the biggest AI companies launched always-on assistants within a few weeks of each other: xAI’s Grok Bot in August, then Meta’s Muse and OpenAI’s Dots in September. When rivals move this close together, it is a strong sign of where the industry is heading.
The Better AI Gets, the Simpler the Screen Becomes
As someone who tracks this field, I see transformations in how we work with AI. A few years ago, we would ask a question in a chat window, then copy the answer into an email or presentation ourselves. Next, popular tools like Claude Cowork and ChatGPT Desktop began building presentations from scratch and editing files while we watched and approved each step.
More recently, we can use those tools to build several agents that could work in the background at the same time, with us reviewing the finished work.
Now, an always-on assistant feels more like a colleague you text: give it a job, leave, and come back when it needs your decision.
What makes these assistants interesting is that they live in a chat, so using one feels as simple as using LINE. Each one also runs on its own computer somewhere else, which is what keeps it “always-on” without requiring you to turn your computer on while it is doing the job.
Your work moves away from learning another tool and towards deciding what to ask for, then checking the result. The assistant can ask permission when it reaches a locked door, just as a colleague would. You remain the person who decides whether to open it.
Useful Assistants Need Careful Access to Company Information
For these assistants to be helpful, they need access to your email, calendar, and files. Giving that access requires care because the assistant is acting inside the systems your organisation relies on.
In China, authorities warned state agencies and banks against installing a popular free assistant of this kind on work devices.
Start by letting the assistant look without taking action. Keep spending limits small, and have a person approve anything that leaves the company. Widen its access only after it has earned your trust through good work.
Managing AI Starts to Look Like Managing a Team
The next AI skill for business people looks a lot like managing a team.
A sales assistant can pick prospects from your list, research them overnight, draft the messages and wait for your yes. A chief of staff can prepare a one-page brief before each meeting, using your emails and past notes.
An inbox assistant can sort new messages, answer the routine ones and flag what needs you. An admin assistant can collect receipts, chase the missing ones and file the claim. A market watcher can check competitor prices or tender notices each day and message you when something changes.
Give the assistant a clear task, the access it needs and a person who checks the work. Organisations should set the rules now, before staff bring their own assistants into the workplace.
Soon your customers will have assistants too, and they will be the ones contacting you.
Oravee Smithiphol is Tech Intelligence & Insights Manager at SCB 10X, where she tracks emerging AI developments across startups, enterprises, and financial services organizations worldwide.
