Just a few years ago, collaborating with AI was a solitary task, confined to a chat tab. It was a 'one person, one tool' structure where a user would ask a question, the AI would answer, and the user would follow up.

However, Anthropic now declares this approach a relic of the past. In a post titled 'Building Effective Human-Agent Teams' published on its official blog on the 24th, the company highlights how the paradigm of collaboration is shifting as AI takes on complex, sophisticated tasks like coding, research, and financial analysis. The way we work is evolving from the era of the 'single player' to that of the 'multiplayer.' The days of a single human unilaterally commanding one agent are over. The future lies in building a 'complete team' where humans set the strategy and multiple AI agents divide the execution.

This article is a report summarizing four key lessons Anthropic learned from months of internal experimentation. Interestingly, these lessons are not entirely new concepts. Clear goal setting, defined roles, thorough documentation, and shared quality standards are exactly the same 'good team habits' that human organizations have adhered to for decades. The difference is that with advanced AI agents joining the team, strictly following these basic rules has become more critical than ever.

From a 'Tool' You Call to a 'Colleague' Who Resides

What does it actually look like to 'work as a team with AI'? Anthropic defines it as a 'multiplayer agent'—an AI that collaborates with the entire organization simultaneously, rather than just an individual.

There is a clear distinction between these and the general chatbots we commonly use. Technically, they may seem similar in that they possess their own memory and skills. However, the decisive difference lies in their 'workspace' and 'qualifications.' A multi-player agent is granted its own credentials, independent of any specific individual. Above all, it resides in the space where actual work takes place. In Anthropic's case, that stage is a collaboration tool like Slack. It is not a one-off tool that only turns on when a user calls it, but rather more like a colleague that is always present in the team channel.

For an AI agent to pull its weight as a member of an organization, three qualifications—akin to a human's 'hiring requirements'—are essential.

First is 'persistent memory,' which remembers team goals and coordinates execution. Second is 'credentials separate from humans,' which ensures the agent operates only within safe and predictable boundaries. Because the agent does not borrow a human's account, the scope of its tasks is clearly defined. Third is 'broad information access,' allowing it to learn how the organization operates and process tasks according to its goals.

Only when these three pieces of infrastructure are in place can an agent join the team. However, joining and performing are two different things; these are merely technical foundations. For a human-agent team to actually deliver results, human change must come first. New ways of collaborating and clear norms that all team members must follow are required. This is why the four lessons derived from Anthropic's internal experiments are drawing attention.

Lesson #1: Work in Public

For AI agents, there is no such thing as 'private conversation' or 'tacit agreement.' Agents understand context only based on text that the organization has made searchable—Slack messages, source code, shared documents, and meeting minutes. Personal direct messages (DMs), water-cooler chat, or documents with restricted access provide no information to the agent. If it isn't recorded and accessible, it doesn't exist to the agent.

For this reason, Anthropic does not set access permissions for individual documents or channels one by one. Instead, it has chosen to set broad, clear 'security boundaries' at the level of the entire Slack workspace or document library. Within these designated boundaries, all context flows seamlessly to both humans and AI. Worrying about whether to make a channel public or share a document on a case-by-case basis causes severe decision fatigue for both humans and AI. Simplifying these boundaries eliminates such inefficiencies at once.

The effect of information transparency is clear. An agent that understands the details of a meeting in real-time will not make the mistake of proposing a task that has already been scrapped. An agent that has learned the product specifications of other teams may even recommend proven success patterns. In particular, because agents read vast amounts of text at speeds far exceeding humans, they act as problem-solvers who frequently point out related tasks that practitioners might have missed.

Of course, this does not mean making extremely sensitive conversations public. If security is required, you can always send a DM to Claude or use existing services (#2, Claude Cowork). Making information sharing the organization's 'default' certainly requires a cultural shift in the workplace. However, the productivity gap between teams that operate with agents holding all the context and those that do not will become impossible to ignore.

Lesson #3: Define Roles Clearly

The key to successful collaboration is not technology, but 'who is responsible for what.' A team combining humans and agents shares one team roster, one set of outputs, and one workspace. In this setup, agents perform different roles, each with their own unique permissions, skills, and tool access. If one agent is dedicated to project data analysis, another might audit design standards, while a third synthesizes research materials.

When a new project begins, team members first communicate with agents to establish specific roles and collaboration processes. Once roles are confirmed, agents work organically, calling upon one another and handing off specific tasks to the right person with the appropriate memory and access rights. The important point here is to provide them with the perfect tools for the job. A data analysis agent needs 'BigQuery' access, and a QA agent needs 'Playwright MCP' to perform at its best.

If role division is blurry, members end up running their own individual AIs separately, leading to duplicated work and fragmented team context. A prime example is tracking data metrics. If one multiplayer agent is dedicated to this task, the entire team can align while looking at the same numbers. Humans should collaborate with agents in the same thread, focusing only on 'final judgment and strategic decision-making'—the domain that only humans can handle.

In fact, one engineering team at Anthropic built a work 'roster' to formalize the roles of humans and agents. By pre-defining agent roles in the form of skill files, they not only made specialization easier but also allowed anyone in the company to quickly replicate and deploy the type of agent they needed. As a project grows, one can simply add an agent to handle new areas. This team recently 'hired' a 'Release Manager Agent' to handle the new software deployment process.

Lesson #4: Set a Clear Goal (North Star)

The decisive difference between an agent that mechanically processes assigned tasks and one that proactively suggests new work lies in 'clear direction.' The most pivotal agents at Anthropic do not stop at completing assigned tasks; they go a step further to propose new projects and workflows. This proactivity emerges when a team that already has rich context (Lesson #5) and clear role division (Lesson #6) adds one final element: an unwavering value-oriented point, the 'North Star' metric.

A North Star is a broad, ambitious, macroscopic goal that helps team members judge for themselves whether the task at hand is truly on the right track. At Anthropic, the entity that establishes this North Star is always human, and it is deeply rooted in the company's mission and business goals. Once the North Star is codified in text, human team members share it immediately with the agents. Humans then make the final decision on which agents to grant the authority to proactively propose new workflows, as not all agents in the team possess the necessary skills and reliability.

There is a real-world example. A team at Anthropic developing internal tools set a North Star to 'make the product onboarding process more useful.' An agent on the team then proactively suggested modifying the text of error messages occurring during onboarding. This small change led to an actual increase in the weekly onboarding success rate the following year. A clearly defined North Star provides agents with consistent behavioral guidelines and opens up practical opportunities for them to actively contribute to team productivity.

Lesson #7: Expand Autonomy in Proportion to 'Trust'

They did not start by handing over 500 bugs to an agent all at once. Anthropic's internal teams grant autonomy only to the extent of the reliability the agent has verified, and then carefully expand the scope of work. While there are success stories where engineers entrusted an agent to independently handle 500 bugs, the start was strictly phased.

Just as it takes time for a new colleague to join a team, gauge their capabilities, and get in sync, the same applies to AI agents. An adaptation period is essential to assign various tasks, understand the agent's performance, and learn how to explain goals and which skill files and prompts yield the desired results. It is also important to re-test existing tasks whenever Large Language Models (LLMs) are updated. This is because the safety guardrails that once prevented an agent from going off-track can become shackles that hinder creative problem-solving for a more advanced, modern model.

The keyword Anthropic emphasizes most in this process is 'verification.' In the long run, the agents that achieved the best results were those equipped with multiple devices to verify their own output before human review. They attach automated tests to source code and apply clear rubrics and style guides to technical documents. When humans establish standards in advance and design every process to be verifiable, the final quality never goes off track. This is why the 'Doer-Verifier' structure, where one agent performs the work and another cross-verifies it, is frequently used.

An engineering leader at Anthropic once formed a task force combining a few human team members and many agents to resolve a massive project backlog. First, a group of agents scanned the entire backlog to check for owners, calculating the complexity of unassigned items and assigning scores. Then, another group of agents selected low-to-medium difficulty tasks from that list and modified the source code directly.

In the initial stage, the human leader reviewed every decision made by the agents to categorize issues requiring human judgment. Later, they trained the agents to judge these exceptional situations and escalate (report) them directly to humans. They built an efficient system where humans intervene only in key decisions involving sophisticated trade-offs.

Furthermore, this team had agents write their own weekly reports documenting 'Lessons & Missteps' to prevent the repetition of the same mistakes. Over time, the leader was able to delegate increasingly complex tasks to the agents, drastically reducing the time spent on routine guidance. Once the agents achieved independent autonomy, they were conversely taught that 'human attention is the scarcest resource.' The agents' behavioral patterns were designed to batch questions, summarize key context to help human team members catch up quickly, and limit the number of items humans review at once.

In the end, it comes back to the fundamentals of a good team

Concluding the article, Anthropic proposes five self-diagnostic kits for human-agent teams.

The destination of these questions is clear: direction, roles, documentation, standards, and the room to learn from mistakes. These are the fundamentals of organizational culture we have long known. AI agents are not some fantasy-like new technology. Rather, they are a mirror that clearly shows how a team falls apart when it skips the basics. The teams that achieve the best results by making a powerful partner like an agent their ally are none other than those that most persistently adhere to these obvious fundamentals.