Now You Can Tag an AI Teammate Into Your Work Chat

On 23 June 2026, Anthropic introduced Claude Tag, a way to bring its AI assistant into a team's chat as something close to a coworker. You tag @Claude in a Slack channel, describe a job in plain language, and it breaks the work into stages and gets on with it, posting back in the thread as it goes. It is in beta for Claude Enterprise and Team customers.

That one sentence hides the actual shift. For most small businesses, AI so far has meant opening a separate chatbot, typing a question, and copying the answer back into the work you were already doing. This turns that around. The assistant sits where the work happens, in the channel where your team talks, and it picks up the task rather than handing you more homework.

It is worth being precise about what is new and what is not. Tagging a bot in a chat is old. An assistant that reads the channel's history, holds the context, works across hours or days on its own, and follows up without being prodded is the part that has changed.

What Anthropic actually launched

Per its 23 June announcement, Anthropic describes Claude Tag as an assistant you tag with a request in simple terms, which then breaks the task into stages and works through them in turn. It is built for a whole channel rather than one person: one Claude runs per channel, so teammates can hand off to it and to each other. It builds understanding from the channel's conversation and the data sources it is connected to, can schedule and run work asynchronously across hours or days, and will proactively flag relevant information and follow up on loose threads. It reaches only the tools and data an administrator scopes to each channel, with strict separation between different uses, and it runs on Anthropic's Opus 4.8 model.

Anthropic points to its own teams as the proof: staff tag Claude to chase down product metrics, work through support tickets, and hunt for the root cause of tricky bugs, and it now writes the majority of the product team's code. For now it is beta and limited to the Enterprise and Team tiers, so an individual Australian small business cannot simply switch it on this week. The direction, though, is unmistakable, and these tiers are usually the front of a queue, not the end of one.

Why this matters for a small team

The promise here is delegation without hiring. When you are the owner, the salesperson, the bookkeeper and the support desk all at once, the bottleneck is rarely ideas, it is hands. An assistant that can be handed a repeatable, well-defined job and trusted to work through it is the kind of help a small business usually cannot afford. This is the move toward agents that Andrej Karpathy and others have been describing for a while: software that takes a goal and works through the steps itself, rather than waiting for the next click.

The value is not the novelty of talking to a bot. It is that the slow, repetitive chase-up and pull-together work, the part that eats an afternoon and never feels finished, can be handed off and watched rather than done by hand. That is real time back, and for a small team time is the scarcest thing there is.

Where the advantage actually sits

Here is the catch that does not make the launch page. An agent with access to your tools and your data is only as good, and only as safe, as the boundaries you give it. What can it touch, what needs a human to sign off, which channel sees which information, what happens when it is confidently wrong. Set up loosely, an eager assistant with the keys is a liability. Set up well, it is a tireless junior who never drops a task. We have written before about the guardrails an AI agent needs before you trust it with real work, and this launch makes that question urgent rather than academic.

So the advantage is not in being first to flip the switch. It is in the design underneath: choosing the right jobs to delegate, wiring in the right context, scoping the permissions tightly, and keeping a person in the loop who can tell useful work from confident nonsense. That is craft, and it is exactly the part that decides whether one of these tools saves you real hours or quietly creates a new mess to clean up. Here is what good looks like once an AI teammate is part of how you work:

• The right jobs handed over: the repeatable, well-defined chase-up and pull-together work, with the judgement calls and the customer-facing decisions kept firmly with people.
• Permissions scoped tightly, so the assistant reaches only the data and tools a given task needs, and anything that spends money or touches a customer waits for a human to sign off.
• Context wired in properly, so it answers from your real numbers and systems rather than plausible-sounding guesses.
• A person who reviews what it produces and can spot the difference between genuinely useful work and an answer that is confidently wrong.
• The whole thing joined up with the rest of how you operate, so it saves real hours week after week instead of just demonstrating well once.

An AI you can tag into the conversation is only as good as the boundaries and the brief you give it. Get those right and it works like a tireless junior. Get them wrong and it is a confident liability holding the keys.NextAura

Treat the exact availability as a moving picture: this is beta, limited to the Enterprise and Team tiers today, and the precise tools and controls will keep changing. What is not in doubt is the shape of the next year, where handing real, multi-step work to an AI inside the tools you already use becomes ordinary. The businesses that benefit first will be the ones who have already decided what they would trust an AI teammate with, and how, before the button is there to press.

This is the work we do at NextAura. We design and build AI agents for Australian small businesses, deciding what to delegate, scoping what they can touch, wiring in the right context, and keeping a human in the loop, so you get the extra set of hands without handing your business to a machine that was never told where the edges are. If you would like to put an AI teammate to work without the risk, get in touch and we will take it from here.