“Hey AI, write me a marketing email.”
“AI, summarize this document.”
“AI, generate some code for this feature.”
This is the assistant paradigm. You ask, it responds. You command, it executes. The AI waits passively until you give it something to do.
And it works. For many things, it works well.
Assistants Have Their Place
Let’s be clear: AI assistants are genuinely useful.
A developer debugging code at 2am doesn’t need an autonomous team. They need a capable assistant that can explain an error message, suggest a fix, or generate a quick utility function.
Someone researching a topic benefits from an assistant that can summarize documents, clarify complex concepts, or answer questions ranging from simple to deeply technical.
For individual tasks - coding help, writing drafts, answering questions, brainstorming ideas - the assistant model is efficient. One person, one AI, direct interaction. No overhead.
This works well alongside kanman.ai too. Use your favorite assistant for individual tasks while your virtual teammates handle the collaborative work. @mention a virtual teammate when you need them involved. Assign tasks when something needs to happen. The systems complement each other.
But Assistants Have Limits
The assistant paradigm hits a ceiling when you try to scale beyond individual, straightforward problems.
Assistants put the cognitive load on you. You have to remember to ask. You have to know what to ask for. You have to phrase requests correctly. You have to review output, iterate, refine, prompt again.
The assistant never takes initiative. It doesn’t notice the project is behind schedule. It doesn’t flag ambiguous requirements. It doesn’t coordinate with other work happening in parallel.
For ongoing work, for projects with multiple moving pieces, for anything requiring coordination across tasks or people - the assistant model creates friction instead of reducing it.
What Teammates Do Differently
Think about human teammates. Good ones, anyway.
They don’t wait to be told every move. They have context about the project. They know what’s coming next. They pull work when they have capacity. They raise issues before they become problems.
A teammate doesn’t need micromanagement. You align on goals, divide responsibilities, check in periodically. The work happens in between.
Virtual teammates work the same way.
In kanman.ai, virtual teammates join your workspace like any other team member. They appear in your board. They can be assigned tasks. They respond to @mentions. They show up in activity feeds.
The interface isn’t a chat window. It’s the same kanban board you use for human collaboration. Same card layouts. Same workflows. Same drag-and-drop prioritization.
Same Boards, Same Tools
This design choice matters more than it might seem.
When AI works through a different interface than humans, you end up context-switching constantly. Chat here for AI, board there for humans, back and forth, mental overhead accumulating.
When AI works through the same interface, the skills transfer. You already know how to assign a task. You already know how to set priorities. You already know how to review completed work.
Virtual teammates slot into workflows you’ve already built. No new tools to learn. No separate systems to manage. Just more capacity on your existing board.
Autonomy Isn’t All or Nothing
“But I don’t want AI making decisions without me.”
Fair. Neither do I, for most things.
That’s why autonomy is configurable. Not a single setting - four distinct levels that match different situations.
Assigned only is assistant mode, basically. Virtual teammates do exactly what you tell them. Nothing more. You’re in full control.
Capacity pull lets teammates grab available tasks when they finish. You still decide what goes in the queue. They just don’t sit idle waiting for you to explicitly assign every item.
Proactive review adds suggestions. Teammates flag potential issues, propose improvements, offer alternatives. You approve before anything changes.
Full autonomy is for mature workflows with clear guardrails. Define the outcomes, set the boundaries, let the team operate. Check in on results instead of supervising every step.
Most people start cautious and adjust. Some tasks stay at assigned-only forever. Some graduate to full autonomy once trust is established. The point is having options.
Real-Time Collaboration
Teammates don’t disappear between requests. They’re present.
In kanman.ai, virtual teammates collaborate in real-time. They edit notes alongside you. They comment on tasks. They update status as work progresses.
This matters for the same reason it matters with human teammates. Visibility creates alignment. You can see what’s happening without asking. Progress doesn’t disappear into a black box between your prompts.
Activity feeds show everything. Human edits, virtual edits, same timeline. You don’t have to wonder who changed what or when.
Bring Your Own Keys (Or Don’t)
kanman.ai works out of the box with built-in AI. No configuration required. Start using virtual teammates immediately.
But if you prefer your own providers, BYOK is available. Connect OpenAI, Anthropic, or Google. Use credits you already pay for. Your keys are encrypted and never exposed.
This isn’t about lock-in avoidance - though that’s nice too. It’s about flexibility. Some teams have compliance requirements about which AI providers they can use. Some have existing contracts they want to leverage. Some just prefer specific models for specific tasks.
Built-in works for most. BYOK exists for those who need it.
The Shift Worth Making
Assistants were the first generation of AI tools. They proved the technology worked. They demonstrated possibilities.
Teammates are the next generation. Not just reactive capability, but proactive collaboration. Not just answering when asked, but participating in ongoing work.
The shift isn’t automatic. It requires rethinking how AI fits into your workflow. But once you experience virtual teammates pulling work, updating status, collaborating in real-time - the assistant paradigm feels limiting.
kanman.ai built around teammates from the start. Not as an afterthought. Not as a feature bolted onto an existing product. The whole architecture assumes AI and humans work together, on the same boards, with the same tools.
Ready to move beyond assistants? Virtual teammates collaborate on the same boards as your human team. kanman.ai - annual workspace subscriptions. €4 / month for individuals, €10 per seat / month.
Marco Kerwitz
Founder of kanman.ai