How we expect ETUNIM AI Studio to work
Four steps from request to result: choosing tools, connecting agents, customising the workflow and following the run. A product still being designed.
Related project: ETUNIM AI Studio
ETUNIM AI Studio is at the concept stage. This article describes what we intend to build, not an available product. The details may change as the project moves forward.
1. Choose your tools
Everything starts with the task. Users choose the AI agents and tools suited to what they want to do, instead of adapting the task to a single tool.
2. Connect your agents
Next, they define how the agents communicate and what information they share. This is where a set of isolated agents becomes a coordinated workflow, able to divide tasks and to recognise and correct errors.
3. Customise your workflow
Users set the instructions, task sequence, priorities, conditions and expected results. Code your logic. Connect your agents. The exact way of doing this is still being defined.
4. Run and optimise
Finally, they run the workflow, follow the results and adjust the configuration whenever needed.
What happens during a run
Automating should not mean losing visibility. We want it to be possible to:
- see the data passed between steps;
- identify the step where an error occurred;
- set human approvals before important actions;
- follow the work from a phone, not only from a computer.
An illustrative example
Imagine a weekly report. One agent gathers information from the chosen sources, another summarises it and a third compares the summary with the defined rules and flags inconsistencies. Before anything is sent, the workflow stops and waits for a person to approve it.
This example shows the idea. It is not a demonstration or the output of a real product, because the product does not exist yet.
How to follow along
See the project page for more detail. We will publish the next steps here, with no dates until they are confirmed.