# Build: Agents that take on real work.

We build agent systems around how your business actually runs, prove them on your real tasks, and hand them over so your team can run and extend them.

## Right for you if

- You know what work you want AI to take on.
- You need it running in your real tools, not a demo.
- You want your team to own it after launch.

## What you get

- **Working agent systems**: Agents connected to your tools, data, and people, within the scope we agree on.
- **Evaluations**: Test sets built from your real tasks, and the bar each system has to meet before it ships.
- **Handover**: Documentation, training, and ownership of the code, prompts, and configuration.
- **Measured results**: A before-and-after comparison against your baseline over an agreed period.

## How it works

1. **Define done**: Agree on the work, the baseline, and the bar each system must meet.
2. **Build and test**: Build with your team and test against your real tasks.
3. **Launch**: Go live with people reviewing the decisions that matter.
4. **Hand over and extend**: Train your team, document everything, and add more work once it holds up.

## Why it works

- **Proven before launch**: Each system has to meet an agreed bar on your real tasks before it goes live.
- **Yours to keep**: You own the code, prompts, configuration, and tests built for you.
- **Built to grow**: Start where the value is clearest, then extend the same foundations to more of the work.

## Questions

**Can you build across several processes?**

Yes. We usually start where the value is clearest, then extend the same foundations, tests, and patterns to more of the work.

**Which AI models do you use?**

Whichever does each task best at the right cost. We test options on your work before choosing.

**Who owns what you build?**

You do. The code, prompts, configuration, tests, and documentation built for you are yours.

**What if a system doesn't meet the bar?**

We tell you, and recommend what to do next. That can include stopping.

## Next step

[Start a conversation about build](https://dag-ai.com/inquiry?interest=build)