Skip to content

Agentic workforce

AI agents are workers now. They take on real work, carry real cost, and need an owner, a budget, and a plan. The Agentic workforce section lets you resource AI agents the same way you resource people. Each agent is a line in your plan, with a forecast, an accountable owner, and actuals that reconcile against it.

Two lenses

Flowstate splits your AI footprint into two lenses:

LensWhat it coversWhere the cost lands
Human teamPeople using copilots (Claude Code, Cursor, GitHub Copilot, Gemini)On the person. Copilot spend is part of their fully-loaded cost.
Agentic teamAutonomous agents doing work with no person at the keyboardOn the agent. Each agent has its own cost line.

A copilot is a tool, never a worker. An agent is a worker, never a tool. The AI cost model explains the split and why nothing gets double-counted.

Where it lives

The section is at Resourcing → Agentic workforce (/plan/main/resourcing/agents), with four tabs:

TabWhat it does
AgentsThe resourced agent list. Create agents, assign owners, deploy them to projects.
AI toolsThe equipment layer: every AI tool your org uses, split into connected providers and services detected through the proxy.
ModelsThe model catalogue. Names, tiers, context windows, and per-token prices.
RulesAdmin only. The policy rules that raise signals against AI traffic.

Results and spend readouts live at Insights → Agent insights (/plan/main/insights/agents), with tabs Dashboard · Spend insights · Agent sessions · Alerts. AI budget caps live at Finance → Budgets → AI caps.

Pages in this section

PageUse it to
AgentsCreate, own, and deploy planned AI workers. The agentic analogue of a headcount line.
ModelsBrowse the model catalogue that prices every agent.
The AI cost modelUnderstand how copilot spend, agent spend, and people costs fit together.
AI service accountsGive each production AI workload its own identity and credentials.
Detected servicesSee every AI service Flowstate has observed traffic for in your org.

How the pieces connect

  1. You create an Agent on the Agents tab and give it an owner and a model.
  2. You deploy it to projects with a run cadence and a monthly spend cap.
  3. You link it to its AI service account. Actual sessions and spend now reconcile against the forecast, plan versus actual, exactly like a person's planned versus actual cost.
  4. Spend rolls up in Agent insights and the forecast, alongside your people. Breaching a cap raises an alert.

Your own usage as a person lives on the personal My AI surface, not here.

Flowstate Documentation