Viktor: The AI Teammate That Fixes Claude’s Team Problem

Picture this. You’re deep in a great Claude session, the draft is sharp, and the idea is finally taking shape. Then you need a teammate’s input. So you copy the output into Slack, explain the background, wait for a reply, and carry the notes back to Claude. Somewhere along the way, you’ve become a very expensive courier.

That’s why this LinkedIn post stopped me mid-scroll. The author calls themselves an “AI freak,” yet says they use less AI than ever before. They open only four apps every day, and one of them is there to fix that exact courier problem.

• Claude (and Claude Code): the main AI workspace
• Slack: for talking to the team
• Wispr Flow: for meeting notes and voice dictation
• Viktor: an AI employee that lives in Slack

I love how short that list is. Most AI stacks keep growing until nobody remembers what half the tools do. This one shrank, and the tool that made the cut earned its place by solving one specific annoyance.

The one thing Claude couldn’t do

The original poster doesn’t complain about Claude’s quality. The frustration is about who gets to be in the room. In their words, Claude is only between them and the AI, so they can’t bring their team into the conversation.

Think of it like hiring a brilliant consultant who’ll only speak to you, alone, behind a closed door. The advice might be excellent, but everyone else on your team only hears your summary of it. Every question from a colleague means another trip back through that door.

So the author did something that sounds a little crazy. They officially “hired” Viktor and added him to the team Slack.

Now the team talks to each other in Slack as usual, and Viktor collaborates with everyone there. He uses both the author’s Claude and their teammate Anisha’s Claude, and he’s connected to all of their apps. The AI stops being a private chat window and becomes a coworker in the channel.

Two people, one AI, one thread

The best part of the post is the walkthrough of how this played out on an ordinary content task. The job was turning the author’s latest newsletter, “Claude for legal,” into a LinkedIn post. Here’s how it went:

1. The author asked Anisha to turn the newsletter into a LinkedIn post.
2. Anisha sent a Slack message tagging @Viktor, asking him to build a page in Claude Code with prompts from the article.
3. Viktor replied with a live page link in 2 minutes.
4. The author jumped in and said they should do it differently.
5. Viktor started working with both of them in the thread.
6. He went to the author’s Substack, pulled the content, built a lead magnet, and replied “Done” with a link.

Step 4 is the one I keep coming back to. In a normal setup, that’s where everything falls apart. The first version isn’t what the boss wanted, so someone has to relay the feedback, reopen a private session, and explain the context all over again.

Here, the disagreement happened right where the work was happening. Viktor could see the request, the first attempt, and the change of direction, so nobody had to play messenger. He just kept going and came back with a finished link.

The creator also compares this to Claude Tag, and they don’t hold back:

“This is what I wanted out of Claude Tag, a proactive collaboration. But it has too many bugs, so I hired Viktor instead.”

That’s one person’s experience, so your mileage may vary. Still, it points at something I find useful when judging any new AI tool. The question isn’t only how smart the model is. It’s whether the model can sit where your team already works together.

How to set it up for your team

The setup this LinkedIn creator shares is refreshingly short:

1. Go to Viktor’s site and add it to your Slack (it works in Microsoft Teams too).
2. Connect the tools you want it to use. The author added the Claude API so Viktor could build in Claude Code.
3. Tag @Viktor in a channel with any task, written the way you’d message a team member.
4. Correct it in the thread when something’s off.

Those last two steps are where the mindset shift lives. You’re not writing a carefully engineered prompt in a separate app. You’re sending the same kind of message you’d send a colleague, then giving feedback where everybody can see it.

A few ways I’d try this

The post doesn’t go beyond the one example, so these are my own ideas for a first test. I’d start with something small and low-stakes, so a wrong turn costs you nothing but a few minutes:

• Content repurposing: “@Viktor, turn the prompts from this article into a simple web page and send me the link.”
• Team feedback loops: have one person request a first draft, then let a second person redirect it in the same thread.
• Quick internal pages: ask for a page that collects the key points from a document, so the whole channel can review one link.

I’d also connect only the tools a task needs at first. An AI teammate plugged into all your apps can do a lot, so it makes sense to widen its access as you learn how it behaves. Keep a human checking anything that goes out publicly, too.

What sticks with me most is how little drama there is in this workflow. Nobody opened a new dashboard or learned a new system. Someone tagged a coworker in Slack, the coworker happened to be AI, and two minutes later there was a link. No courier required.

Check out the full LinkedIn post to watch the video of this thread unfolding. Seeing the back-and-forth live makes the idea click much faster than any description can.