The video above shows the whole thing happening: a workflow described in plain English, AI generating real Python code, and the result published for production. Watch it first, then follow the steps below to build your own. No coding experience needed.
What Makes This Different
Three things to notice as you watch:
- The code is visible. Build Studio doesn't hide what it builds behind a spinner. It generates Python you can open, read, and review. That's the difference between trusting a tool and auditing one.
- The preview runs against real data. You don't approve a description of what the workflow will do; you approve actual output from your actual sample file.
- Publishing freezes the code. Once the output looks right, the workflow becomes a versioned, immutable artifact. No AI runs at execution time: same input, same output, every time. That's what compliance teams and auditors need to hear.
What You'll Build
A workflow that takes a monthly sales Excel file, calculates key metrics (total revenue, top products, regional breakdown), and generates a formatted summary report.
Time required: ~10 minutes
Step 1: Open Build Studio
Log in to Dittah Studio and open Build Studio. You'll see your workflows list; click New Workflow to start fresh.
The editor walks you through two steps: Define, where you describe your inputs and instructions, and Generate & Test, where the AI writes the code and runs it against your data.
Step 2: Describe Your Input
In the prompt area, describe your data source:
"I have a monthly sales report in Excel format. Each row has: date, product name, region, quantity sold, and unit price."
Dittah will ask clarifying questions if needed, or directly generate the ingestion step. You can also upload a sample file so the AI can inspect the actual structure.
Step 3: Describe the Transformation
Now tell Dittah what you want to calculate:
"Calculate total revenue per product and per region. Find the top 5 products by revenue. Show month-over-month growth percentage."
The AI generates Python code that performs these calculations. You can see the generated code, but you don't need to understand it. The preview output shows you exactly what the result looks like.
Step 4: Preview and Iterate
Click Continue to Generate & Test to have the AI generate the code and execute it with your sample data. You'll see:
- A table with revenue by product and region
- The top 5 products ranked by revenue
- Growth percentages compared to the prior month
Not quite right? Just refine your description. For example:
"Also exclude any rows where quantity is zero. And round all percentages to one decimal place."
Dittah regenerates the code. Preview again until the output matches your expectations.
Step 5: Add an Output
Decide where the results should go:
"Export the summary as an Excel file with two sheets: one for the regional breakdown and one for the top products. Also send a brief email summary to the team."
Dittah adds output steps for file generation and email delivery. You configure the email recipients in the workflow settings.
Step 6: Publish, and the Code Freezes
When you're satisfied with the output, click Publish Workflow. Publishing freezes the generated Python code into a versioned, immutable artifact. From this point on, no AI is involved in execution. The frozen code runs deterministically: same input, same output, every time.
Once published, the workflow is available to run. You can:
- Run on demand: trigger manually from the Ops Hub
- Schedule it to run daily, weekly, or monthly on a cron schedule
- Trigger via API, calling the workflow from an external system
What Happened Behind the Scenes
Here's what makes this different from other AI tools:
- AI designed the workflow: it understood your intent and generated clean Python code
- You tested and approved it (the preview let you verify before committing)
- The frozen code runs without AI, so there are no LLM calls at runtime, no hallucination risk, no variable outputs
- Every execution is auditable; Ops Hub logs inputs, outputs, and timing for every run
This is the Dittah philosophy: AI builds, code runs.
Next Steps
- Getting Started guide, if you haven't set up your environment yet
- Build Studio deep dive to explore all the capabilities
- From Raw CSV to Professional Video, the other demo: a spreadsheet becomes a narrated video in minutes
- Subscribe on YouTube for more demos as we publish them
