AI Teammates Need Better Context Than Slack Can Provide

Claude Tag is an important signal — AI is moving from “open a separate chatbot and ask a question” to “bring the AI into the place where work is already happening.”

That part makes sense.

Anthropic is starting with Slack, where Claude can join channels, read selected context, connect to tools and data, remember relevant information, and work asynchronously on tasks. In other words, Claude is not just a chatbot anymore. It becomes more like a teammate that can be tagged into a conversation.

This is directionally right.

I’ve been thinking about the same problem while building VOSTORQ, where AI agents were designed to be participants in the conversation from day one. AI becomes much more useful when it is not isolated in a separate browser tab. Work is collaborative. Decisions are collaborative. Context is usually created between people, not inside a single prompt.

So yes, AI should be part of team communication. But with Slack, there is a big caveat. Slack is still a chat. It always was, and it always will be.

Convenience became infrastructure

Slack and similar messengers became the main communication systems in many companies because they were convenient. Too convenient, maybe.

They are fast. They are easy. They feel lightweight. You can create a channel in seconds, send a message instantly, loop someone in, start a thread, drop a file, react with an emoji, and move on.

That convenience is exactly why they spread everywhere. But convenience does not automatically make something a good foundation for company communication in the long run.

Chat is good for quick coordination. It is useful for short questions, small updates, and informal team awareness. The problem starts when companies treat chat as the place for everything: decisions, requirements, product history, technical discussions, customer context, project memory, process changes, ownership, approvals, and long-running work.

That is where chat starts to break down.

The mess

Real work conversations are messy.

People jump between topics. Someone answers a question from yesterday. Someone else references a call that half the channel did not attend. A decision happens in a thread. The reason behind that decision is buried 30 messages above it. The actual constraint is in a document nobody linked. The customer context is in a CRM, a support ticket, or someone’s memory.

Then a few months later, everyone is trying to reconstruct why something happened. This is already difficult for humans. Now Claude Tag has been invited into that same mess.

That can still be useful. Claude can read more, remember more, and connect more tools than any human reasonably can. It can follow channels, summarize threads, pull in data, and work in the background.

But if the communication is unstructured, the AI is not escaping the problem. It is working inside the same problem.

AI does not remove the need for structure

There is a common assumption behind many AI features right now:

AI can read all of this messy context and figure it out.

Sometimes it can. But that is not the same as having a good system.

If the underlying communication is scattered, informal, and constantly changing, the AI has to spend a lot of its effort rebuilding context that should have been captured properly in the first place.

What was the actual decision? Who owns the next step? Was that suggestion accepted or rejected? Is this message a final answer or just someone thinking out loud? Which part of this thread matters three months from now? Was this conversation about a temporary issue, a product rule, a customer-specific exception, or a company-wide policy?

Humans struggle with these questions in chat. AI will struggle too, especially when parts of the context are outside the channel, hidden in private conversations, sitting in meetings, or never written down at all.

Adding AI to Slack may make the mess more searchable and more responsive. It does not clean it up.

The real problem is not where the AI sits

Claude Tag is interesting because it shows where things are going — AI needs to participate in team communication. It needs to understand what people are discussing, help with follow-up work, notice unresolved issues, and preserve useful context.

That is the right direction. But the bigger problem is not whether we can tag AI in Slack. There were already a bunch of solutions for that before Claude Tag.

The bigger problem is whether the conversation itself is designed to create usable context. A chat stream is not really built for that. It is built for flow. That is the weakness. Flow is great when the goal is speed. It is not great when the goal is durable knowledge.

Companies do not only need faster answers. They need better memory. They need a way to turn conversations into decisions, decisions into knowledge, and knowledge into something reusable.

The future is not just AI in chat

I do not think Claude Tag is wrong. I think it is a natural and useful step.

For many teams, having Claude inside Slack will be helpful. It will save time. It will reduce some repeated explanations. It will make AI feel more collaborative. It will probably become normal for teams to delegate work to AI agents inside the tools they already use.

But I do not think chat is the final form of AI-assisted company communication. Chat is too unstructured for that. For quick coordination, Slack may be great. For decisions, long-running work, product history, reusable company knowledge, and anything that needs durable context, chat is a weak foundation.

AI teammates need more than access to conversations to be effective — they need conversations structured enough to become useful context. That, I think, is the part that matters most: not just putting AI where people talk, but building communication systems where what people talk about can actually become company knowledge.