Instruct a prompt to record
- Updated: 2024/06/11
Developers can save time and effort by entering instructional prompts in Co-Pilot for Automators to engage the recorder when building bot automation.
Using the recorder
Co-Pilot for Automators is designed to simplify the work of building automation. It supports Recorder functionality, which leverages natural language prompts. This means you can describe your goals and Co-Pilot will generate the vital components of a Recorder, making the development process easy.
Prompt to record is a tool that allows you to automate a wide range of tasks. For instance, you can record a series of tasks in a process or record your interactions in a web browser. These interactions are then translated into actions in an automation, making the process of automation development more efficient.
With the Prompt to record tool, you can accelerate your automation development. You can use natural language prompts to quickly generate actions in the Bot editor when building an automation. With the required actions automatically added to the canvas, you can automate web pages reliably.
Benefits of prompt to record
Co-Pilot for Automators offers a single dialog box where developers can write a set of natural language instructions and send this prompt to generate and orchestrate multiple actions on the canvas. Instructions can include any inputs (ex. URL, text, variables) and describe how to use them with the recorder package (ex. capture, set text). With support from generative AI capabilities, the prompt is interpreted and the automation scenario is generated according to the actions and configurations recognized by Co-Pilot for Automators.
By sending a prompt to record, developers can accelerate the process of building with the recorder package by combining the work of adding and sequencing actions in the canvas into the single task of writing the instruction for the prompt. Rather than spending time and concentration on manual orchestrating actions in the canvas, developers can first focus on outlining the scenario with recorder actions and inputs in the instructions of the prompt. After the outline of these steps has been generated and orchestrated on the canvas, developers can begin specializing the automation to specific needs by setting variables and verifying configurations.
Define instructions of the prompt
When writing the instruction of your prompt in Co-Pilot for Automators, you directly engage the recorder package from the Bot Editor to orchestrate the sequence of actions described in your prompt. The recorder is recognized by the instruction /recorder. Any following instruction is interpreted as how and what to record (ex. /recorder go to URL).
Best practices and verification
When you send a prompt that points to a URL, a tab ID is created and referenced while the Co-Pilot page is open. The Control Room continues to reference the established URL as you continue to build the automation from the same tab. Keep the Co-Pilot page open to continue using the established tab id in subsequent prompts. Only provide a new URL if the bot needs to reference a new website or you have closed the page and are returning to continue development.
To write the most effective prompt, include all the instructions that describe the automation scenario in a single prompt. Since this scenario involves recording across pages and components within a targeted site, you might find it necessary to divide your instructions into all the operations for one page or component at a time. When the scenario progresses across pages or nested components, the best experience is to pop-out the targeted tab in an adjacent window to simultaneously view with the browser window displaying the Bot editor. This experience allows you to accurately view and identify components when instructing the prompt to record.
As actions are generated via prompt, it is important to verify the results added in your canvas and ensure the bot task accomplishes your desired goal. We suggest testing the bot before proceeding any time multiple steps are generated from a single prompt.