Claude Cowork on Mobile and Web: The AI Agent That Works With Your Laptop Closed

On July 7, 2026, Anthropic made its Claude Cowork agent available on the web and on the phone, after it had been limited to a desktop application launched in January. The change sounds minor, and it touches something genuinely new: an AI assistant that keeps working on a task even when your computer is closed, and sends you the result later.

For a business in Trois-Rivières or elsewhere in the Mauricie, the question is not whether Claude Cowork is impressive. It is understanding what it really means to hand a task to software that runs it on its own in the background, and where the limits sit before you let it near your data.

1. What Claude Cowork just added

Claude Cowork is an agent built for office work in the broad sense: assembling a report, reconciling a spreadsheet, drafting a proposal, building a checklist. Until now it had to run in a desktop application, with the computer on and connected. The new capability changes three concrete things:

• Mobile and web access: you can start or follow a task from your phone or a plain browser.
• Background work: the task keeps running even if your device is closed or offline. Before, a stable connection was required for the whole job.
• Follow-up across devices: you start a task at the office, get status updates on your phone, and pick up the finished work later.

The rollout is in beta, first for Max subscribers, then for the rest of the paid plans over the coming weeks. The desktop application remains useful for deep work that touches your local files and your browser.

2. An AI agent is not just a chatbot

The difference between an agent like Claude Cowork and a plain conversational assistant is worth spelling out, because it explains why the news matters. A chatbot answers a question and waits for the next one. An agent receives an objective and works through the steps on its own to reach it: reading documents, filling a spreadsheet, comparing figures, producing a deliverable.

In practice, that means you delegate a task rather than a question. You are not just asking for an answer, you are handing over a ten-minute or one-hour assignment. Being able to launch that assignment from a phone and let it finish unsupervised brings the tool closer to a real digital teammate. It is also what calls for caution, because an agent that acts alone in your data can go wrong alone.

If you want to govern this kind of tool in your company rather than letting each employee use it on their own, that is exactly the sort of support an IT partner like our managed IT services can put in place.

3. What the real usage numbers say

Alongside the mobile launch, Anthropic published data drawn from 1.2 million anonymized Claude Cowork sessions, collected between May 11 and May 31, from more than 600,000 organizations. The picture is instructive, mostly because it breaks a common assumption about AI.

• 33.4% of the work involved business processes: reconciling spreadsheets, assembling reports, the kind of thing you find in finance, human resources and administration.
• 16.4% covered content creation and writing: drafts, presentations, proposals.
• Only 8.7% was software development, even though these agents came out of the programming world.

In other words, most people use an advanced AI agent for perfectly ordinary office work. That is good news for a smaller business: the most realistic time savings are not in a high-tech project, they are in the spreadsheets and reports somebody at your company rebuilds every week.

4. What this changes for a Quebec business

The value for a small or mid-sized company does not come from the model’s name, it comes from what you can hand off without spending the day on it. A few realistic examples we see with clients in the Mauricie and Quebec City:

• Recurring reports: a monthly sales or expense summary built from an export, instead of rebuilding it by hand.
• Spreadsheet cleanup: reconciling two lists, spotting duplicates, standardizing columns.
• First drafts: a proposal, an internal policy, a client email to review afterward.
• Asynchronous work: launching a task before you leave for the evening and collecting the draft the next morning, laptop closed.

The trap would be believing this replaces your employees’ judgment. It does not. An agent produces quickly, but you still have to read it over, check the numbers and own what goes out. The real gain is removing repetitive work to free up time for what needs a human brain.

5. Precautions before letting an AI agent loose in your data

An agent working on its own in the background, on your documents, deserves a framework. Before you open the door to this kind of tool in your business, three basic reflexes:

• What data it can see: avoid giving it access to sensitive information (clients, finances, personal data) without knowing where that data travels and where it is stored. Law 25 still applies.
• Who is allowed to use it: a clear policy beats wild adoption, machine by machine, with personal accounts.
• Always read it over: a deliverable produced by an agent is a draft, not the truth. Human verification stays mandatory, especially on numbers.

These are exactly the questions we help settle before an incident happens. A little structure at the start avoids a lot of headaches, and it is easy to set up with the right guidance.

Frequently asked questions

Is Claude Cowork free?

No. Mobile and web access goes first to Anthropic’s Max subscribers, in beta, then widens to the other paid plans. AI service pricing changes often, so always check the current price on Anthropic’s official site before committing.

What is the difference between Claude Cowork and Microsoft Copilot?

Both are AI assistants, but Claude Cowork is an agent that carries out complete tasks in the background, while Copilot is mostly built directly into the Microsoft 365 applications. The better choice depends on the tools you already use and what you actually need.

Is it safe for a smaller business?

It can be, provided you govern access to your data, define who may use it and always review what the agent produces. The risk is not the tool itself, it is adoption with no rules and no supervision.

In short, a promising tool that deserves a framework

Claude Cowork arriving on mobile and web marks a step: AI is moving from the assistant that answers to the agent that acts, even when you are not in front of the screen. For a Quebec business, the potential is real on repetitive administrative work, provided you keep control of the data and the verification. If you are wondering how to bring this kind of tool in without exposing your company, have a look at our managed IT services or contact our team in Trois-Rivières to talk it through plainly.