Spend Docs
Agents

Create a New Agent

Follow these steps to initialize a new AI model. Creating an Agent involves linking a verified data source to your desired classification structure so the machine learning engine can begin identifying spend patterns.

🛠️ Step-by-Step Setup

  1. Start Creation: When you are on the Agents module page, click on the + Create Agent button. Agents

  2. Enter Agent Info: A dialog box appears. In that enter the Agent Name (e.g., ICP_03) and a Description.

  3. Select Training Data: Select the specific Dataset you want the Agent to train on from the dropdown menu.

  4. Select Taxonomy: Select the Taxonomy that defines the spending categories the Agent will use.

  5. Finalize: Click Create Agent. The Agent is now initiated and begins its training process

  6. Define Identity: In the creation dialog, enter a unique Agent Name (e.g., North_America_Indirect_v1) and a clear Description of its intended scope.

  7. Select the Training Source: From the Dataset dropdown, choose the specific dataset the Agent will use for its initial learning phase.

Note

For the best results, ensure this dataset has already undergone Vendor Normalization.

  1. Map the Target Taxonomy: Select the Taxonomy that contains the category hierarchy you want the Agent to output. This acts as the Agent's "dictionary."

  2. Finalize & Launch: Click Create Agent. The system will immediately begin the Model Initiation phase.

⏳ What Happens Next?

Once you click Create, the Agent moves through two automatic states:

  • Training: The AI is analyzing the textual patterns in your selected Dataset and mapping them to the Taxonomy.
  • Ready to Classify: Once the model reaches this state, you can use it to classify any other compatible dataset across the platform.

💡 Success Criteria

  • Data Quality: Use a training dataset that is as "clean" as possible. The more accurate your initial dataset, the less manual auditing you will have to do later.
  • Taxonomy Alignment: Ensure the Taxonomy you select has enough depth (Level 3 or Level 4) to capture the nuances of the data in your training set.

Next Step: Once your Agent is "Ready," you can begin Auditing and Configuration to refine its accuracy.

On this page