Howdigitalcolleaguesaregoverned
How digital colleagues are governed: real identity, a readable practice score, a cost guard before every call, and a record of every action taken on someone's behalf.
Digital colleagues are full members of the team. That raises a question, usually in the first 10 minutes of a conversation with security and risk: who is accountable, and how would we know? Everything below is something an administrator can read on screen.
Governance is part of how the team works
Much of the market sells human-in-the-loop as a safety feature, which treats the agent as a tool that needs supervising. We take a different view. The team is the unit, and governance is part of how the team works. If a change to production needs two approvals, it needs two approvals whoever proposed it. A colleague follows the same approval step as a person, and passes a matter to a person for the same reasons: it is outside the colleague's job, it is unclear, or the stakes are high. There is no separate set of governance rules to maintain.
Identity
Every colleague has its own chat account and its own client in your identity provider, both created when you save the colleague. The platform works out who is calling from the credential, so a colleague cannot pretend to be another. What it can reach is set before it asks, so a request outside its scope cannot happen.
Practice score
Each colleague is scored continuously against a rulebook of 31 practice rules in seven categories: model choice, tools, knowledge, prompt, safety, cost and lifecycle. Every finding names the rule, the consequence and the fix. The rules are settings, so an administrator can change the weight of a rule or remove it without a software release. You can look at a colleague's score before you put it in a client room.
Cost
A cost guard runs before every call. It checks how many colleagues are chained together, whether one is calling itself in a loop, how much the conversation has already cost, and whether one colleague is allowed to call another at all. Recurring work is capped by the number of runs, and cost findings are shown for each colleague. We offer limits on each call and each colleague, and no single hard limit on monthly spend.
Acting as a person
Sometimes a colleague acts on your behalf, for example booking your leave under your permissions and in your name. It does this only in a direct message, only when your identity is linked, and only after it asks you to confirm. The credential it uses covers that one action and expires in minutes. Every action taken on your behalf is recorded, and an administrator can read the record: who the colleague acted as, which tool it used, whether a person approved it, and what happened. If somebody asks who is accountable, the answer is the named person the colleague acted as, with their confirmation recorded.
Where the data goes
Rooms are encrypted by default, and so are calls. Knowledge is set up for each colleague and fetched only when it is needed. Each message sends a short summary and does not send the room's history. The agent behind a colleague can be one you run yourself: bring your own using the open agent-to-agent protocol, or run one on Amazon Bedrock. Routing, memory, governance and identity stay as they are when you change the model.
Ask to see a practice score, a record of an action taken on someone's behalf, or a cost finding, and we will show you the screen.
See the whole colleague
Governance is one part of what makes a digital colleague a real member of the team. See the full proposition, and how to start with one job in one room.