Skip to content

Implementing Atlas AI agents

Early adopters

The features described in this section are currently available to early adopters only and are subject to change.

Follow these steps to implement Atlas AI agents in your organization.

  • 1⃣ Scope your use case—identify a problem to solve or a workflow to automate with an Atlas AI agent. Define requirements like features, performance, and cost.

  • 2⃣ Identify and prepare your data—identify a representative evaluation data set that allows you to measure the agent's performance against your use case.

  • 3⃣ Choose the language model—the language model library has detailed reference information about the available language models.
  • 4⃣ Build the agent—use the low-code Atlas AI Agent builder to rapidly build and prototype agents.
  • 5⃣ Test the agent—publish agents for testing and monitor performance for further refinement. Conduct red-teaming for potential misuse and A/B test improvements.
  • 6⃣ Deploy to production—once your agent runs smoothly end-to-end, deploy to production.
  • 7⃣ Monitor and improve—monitor performance and effectiveness to make ongoing improvements.

Step 1: Scope your use case

Identify a problem to solve or a workflow to automate with an Atlas AI agent. Define requirements like features, performance, and cost.

Step 2: Identify and prepare your data

Identify a representative evaluation data set that allows you to measure the agent's performance against your use case.

Step 3: Choose the language model

We recommend that you use the information in the Choosing a language model article and the Language model benchmark report to carefully evaluate and find the best language model for your agent and use case. Also, the Language model library has detailed reference information about the available language models to help you choose.

Step 4: Build the agent

Use the low-code Atlas AI Agent builder to rapidly build and prototype agents.

Step 5: Test the agent

Publish agents for testing and monitor performance for further refinement. Conduct red-teaming for potential misuse and A/B test improvements.

Step 6: Deploy to production

Once your agent runs smoothly end-to-end, deploy to production.

Step 7: Monitor and improve

Monitor performance and effectiveness to make ongoing improvements.