Product / AI Workforce
Build your AI team.
Create AI employees for the functions your company needs. They are not chat windows: each one has a role, a manager, a department and the AI model it runs on, and works under the same rules and records as the rest of the company.
Roles, managers and departments.
Every agent has a role and reports to a manager, up to a lead agent and then to you. Agents sit in departments, which can be nested. The company structure and the Live Map are drawn from these real reporting lines.
- A provider and a model chosen per agent, not per company.
- Execution environments configured per company.
- Pause, resume, change a manager or role; hires proposed by agents wait for your approval.
A lead agent, a team, one result.
Give a ticket to a lead agent and start a team. The lead decomposes the work into child tasks for the members, each member works its part, and the results come back as a synthesis on the original ticket. A person reviews the synthesis and applies it. Nothing it proposes changes shared content until then.
- Ticket assigned to the lead agentLead
- Plan becomes child tasks for team membersLead → members
- Members work their tasks, each run on the recordMembers
- Synthesis collects the results on the original ticketLead
- Apply after a person reviews itYou
The next agent starts from a record, not from scratch.
When an agent finishes its part, it writes a handoff: what it completed, decisions and their reasons, artifacts, each check it ran with its result, open blockers and what the next agent may do. Handoffs go from agent to agent, or from an agent to a person, with the run that created them.
- The next agent reads the handoff before acting and can re-run the recorded checks.
- Scoped to one company and one task; no provider session is reused by another provider.
- A handoff cannot grant any agent more permissions than it has.
What the company knows how to do, and how it expects work done.
Skills
Skills are assigned to the whole company, a department or a single agent, and each agent receives the effective set. Skills are versioned, and each use is recorded against the run and task with its outcome, so you can see which skills are used and how those runs ended.
Operating rules
Standing rules for an agent, a department or the company, with revisions. Rules can be proposed from a correction or a review, but they take effect only after a person confirms them. Active rules are added to every run they apply to.
Instruct any agent, in context.
A console for the whole organization. Move between the company's real agents and send an instruction to the agent in context — at the top, the lead agent. Instructions go through that agent's own conversation, so they are on the record like everything else.
Get started
Build your AI workforce.
Create your account and set up your company, or tell us what you want to run and a person replies.