Digitas · Intelligent Creative
Multiplayer AI
When Agents Join the Room
For the past few years, using AI has mostly been a private activity. You open a chat, ask a question and get an answer. Whatever context the AI has stays inside that conversation. The strategist sitting next to you cannot see it. The designer working in the same file cannot build on it. The account lead joining tomorrow may never know it happened.
That was fine when AI was mainly a personal productivity tool. It makes less sense when the work itself is collaborative. Most creative and product work already happens in shared spaces such as Figma, Miro, Slack and Notion. The people are there. The AI usually is not.
Multiplayer AI changes that by moving agents into the tools teams already use. They work from the same context, contribute to the same artefacts and leave behind work that everyone can see, question and develop. More importantly, people and agents can work on the same task at the same time.
From private answers to shared work
Most AI use is still single-player. Someone asks a question, gets an answer and carries it into the actual work. It gets pasted into Slack, dropped into a deck or summarised in a meeting. By that point, some of the context has usually disappeared. The team sees the output, but not always the thinking behind it.
Multiplayer AI starts with a different assumption: the shared workspace is the source of truth, not one person’s chat history. While one person develops an idea, an agent can research the subject. While a designer sketches an experience, another can challenge the proposition. A strategist can see both develop and redirect the work as it happens.
Everyone can see what everyone else is doing, including the agents. That is a much bigger shift than making a chatbot conversation shareable.
When agents enter the canvas
Imagine a FigJam board during an ideation session. The brief is pinned at the top. People are adding notes, grouping themes and developing ideas. At the same time, one agent is pulling useful research, another is responding to ideas and a third is spotting patterns elsewhere on the board.
Their cursors are visible. Their contributions appear beside the human work. The team does not have to stop, open another tool, write a prompt and paste the answer back. People and agents are working together in the same place.
In this kind of workspace, agents take part directly on the canvas. They read the brief from the board, appear through named cursors and add comments beside the team’s work. They might offer provocations at the beginning, react to ideas during a sprint and help turn the strongest thinking into clearer principles at the end. Their contribution is not hidden in a transcript. It becomes part of the board.
A named cursor may sound like a small detail, but it changes the relationship. The team is no longer stepping away to “use AI”. The AI is already in the session.
From turn-taking to working together
Private AI usually creates a relay. A person thinks, asks the AI, receives an answer and decides what to do with it.
Multiplayer AI is less linear. Humans can add ideas while agents research, critique or synthesise. One agent can build on something written moments earlier. Another can challenge an assumption before the workshop has moved on.
The value is not simply that AI produces things quickly. The work becomes faster because the coordination becomes faster. It also becomes more visible. You can see who added what, what they were responding to and how the work developed. That matters if teams are going to trust agents as contributors rather than treat them as a black box.
Building the operating model
The deeper change is not about model quality. It is about how the work is designed.
In many teams today, AI is added after the main activity. A workshop happens, someone runs the output through ChatGPT and the results are pasted back in.
When agents are part of the process from the beginning, people and agents can work from the same brief. Research, provocation, creation and criticism can happen while the work is still developing. By the time the session ends, the shared file already contains the ideas, decisions and disagreements that shaped it.
At Digitas, we are building this operating model through the newly formed Intelligent Creative team. It brings people, specialist agents, shared tools and creative processes into the same working environment.
Multi-agent systems give agents different roles. Multiplayer AI gives them somewhere to work with people.
A better room
We are moving from “I asked the AI” to “the AI was in the session”.
Single-player AI helps an individual move faster. Multiplayer AI helps a group move together.
The next meaningful unit of AI value may not be a better answer sitting inside somebody’s private chat history. It may be a better room, where people and agents can see what each other is doing and build the work together.