Conbase.ai Recommended Workflow (Detailed Guide)
Step 1: Data Preparation
High-quality output starts with clean, well-structured input data. A little preparation upfront ensures the AI has the best possible context to work with.
- Export Required Data: From your source system (e.g., PIM, ERP, Shopify), export only the data fields and rows you need for the task. Avoid exporting unnecessary columns.
- Review and Clean: Upload the CSV to a tool like Google Sheets or Excel. Scan for inconsistencies, typos, or missing values. Ensure data formats are consistent (e.g., all dates are in the same format).
- Upload to Conbase.ai: Once your dataset is clean and focused, upload it to your Conbase.ai project to get started.
Pro Tip: The cleaner your input data, the faster and more accurate your results will be. Time spent here saves significant time in refinement later.
Step 2: Pipeline Configuration
This is where you instruct the AI on what to do. A well-defined prompt is the key to getting the exact output you need.
- Define the Desired Result: Be specific about the format, structure, length, and any domain-specific features you want. For example, "a 150-word product description in a friendly tone."
- Map Required Data Fields: Tell the AI which columns from your CSV to use as input for the generation.
- Add Static Context: Provide any additional static information that is relevant for all rows. For instance, if you're processing products from a specific category, add a detailed category description to the prompt.
- Set Correct Parameters: Ensure your prompt configuration (e.g., processing mode, AI model, integration mode) is correctly set up for your specific use case.
Step 3: Testing and Refining
Before processing your entire dataset, run a small-scale test to validate your prompt and configuration. This iterative process allows you to refine the output without wasting resources.
- Run a Small Batch: Test your prompt on a small sample of rows. For simple tasks, 5-10 rows might be enough.
- Increase Sample Size for Complexity: For more complex or nuanced results, a larger test set of 20-50 rows will provide better insights into the AI's performance across different scenarios.
- Analyze and Refine: Review the generated output. If it's not perfect, adjust your prompt instructions, context, or examples and test again until you are satisfied with the quality.
Step 4: Processing and Exporting
Once you have prepared your data and refined your pipeline, you are ready to process the full dataset.
- Run the Full Pipeline: Execute the generation process on your entire clean dataset.
- Export the Results: After the processing is complete, download the enriched CSV file containing both your original data and the newly generated content.
Step 5: Review Your Results
The final step is to conduct a final quality check before importing the data into your target system.
- Use the Data Viewer: Quickly review your data directly in the Conbase.ai data viewer. The cell viewer can even render basic HTML, which is useful for checking generated descriptions.
- Inspect in a Spreadsheet: For large datasets, it's highly recommended to upload the final CSV to Google Sheets or Excel. This allows you to easily sort, filter, and spot-check the results at scale.
- Import to Target System: Once you are confident in the quality of the data, proceed with importing it into your target system (e.g., Magento, Shopify, PIM).
Start Automating Your Content Workflow
Ready to Build Your First Pipeline?
Transform your workflow from manual and slow to automated and scalable. Sign up and build your first AI content pipeline in minutes.
Get Started with Conbase.ai Today
