---
url: /handbook/opportunities.md
description: >-
  How DutyGraph turns a mapped company into improvement and AI opportunities,
  and what is Built today.
---

# Find opportunities

Once the work is mapped as Task Cards, DutyGraph helps the advisor find what to change. It shows where the hours go, which work could use AI, what may limit a flow, and the cheapest change that can work. A person decides at every step.

## The path, in order

```mermaid
flowchart LR
    TC["Task Cards<br/>in the Company Repo"] --> H["Hours a week<br/>Reported by the team"]
    TC --> F["AI-fit pass<br/>code proposes a mode"]
    F --> A["Proposed agents"]
    H --> A
    TC --> C["Possible constraints<br/>accepted with an owner"]
    C --> R["Route<br/>cheapest change first"]
    R --> O["Outcomes<br/>measure, pilot, result"]
    A --> B["Agents builder"]
    O --> RP["Reports"]
```

| Step | What happens | Who acts | Where in the app |
| --- | --- | --- | --- |
| Hours | Each person taps a range for the hours a week a Task Card takes. Code reads the range. | The employee | Team asks, then **Opportunities › Your team** |
| AI modes | Code proposes an AI mode, a performer and an owner for each Task Card, with a plain reason. | Code proposes. The advisor accepts. | **Repo › To review** |
| Proposed agents | Code proposes an agent for a Task Card with an owner and an AI mode. | Code proposes. The advisor opens it. | **Opportunities › Proposed agents** |
| Possible constraints | AI suggests what may limit a flow, with quotes from Repo files. | The advisor accepts or rejects each one. | **Opportunities › Possible constraints** |
| Route | Code walks a ladder of changes, cheapest first, for each accepted possible constraint. | The advisor chooses the change. | **Opportunities › Changes** |
| Outcomes | The change gets a measure, a prediction and a pilot. Code sets the result word. | The owner records readings. | **Outcomes › Results** and **Outcomes › Measures** |
| Reports | Dated reports for the sponsor, by loop step. | The advisor saves, approves and delivers. | **Reports › Prepare** and **Reports › Saved** |

## Where AI is used, and where it is not

Most of this section is code. Code gives the same answer for the same Repo, and it makes no AI call.

* **Code, no AI:** reading the hours ranges, the AI-fit pass, proposing agents, the constraint signals, every Route verdict, fit and value, the proof rule, and every report.
* **AI, after the advisor consents:** **Find possible constraints**, a framework run (RACI, Theory of Constraints, SWOT / TOWS) and DutyGraph Agent. Each uses the company's approved AI connection. AI only proposes. The advisor accepts or rejects each result.
* **Sample companies:** no AI call at all. Their output says **Sample output · no AI call** or **Proposed by code · sample**.

DutyGraph shows hours, never money. It shows no savings and no return on investment. It does not run a change, an automation or an agent.

## Pages in this section

* [Opportunities](opportunities.md): hours a week, the AI-fit pass, proposed agents and assigning an opportunity to an employee.
* [Outcomes and Strategy](outcomes-and-strategy.md): possible constraints, frameworks, DutyGraph Agent, measures, pilots and the weekly review.
* [Route](route.md): the cheapest change that can work for each accepted possible constraint, and how to prove it.
* [Reports](reports.md): the report catalogue, saved snapshots, delivery and the Company Snapshot.

Next, turn a proposed agent into a draft in [Build agents](../agents/index.md). Then govern its access in [Govern agents](../governance/index.md).

::: info
**Status.** Hours, the AI-fit pass, proposed agents, possible constraints, the three framework exhibits, Route, Outcomes and the Reports tab are **Built**. Hours with an agent are authored sample values on the sample companies only. Money values, a recorded with-agent prediction for a real company and the other framework exhibits are **Planned**.
:::
