How to Analyze the ROI of GUI Agents

In order to scale investments in any solution every enterprise customer needs to understand how to calculate the ROI.

Here’s a method to analyze the ROI of GUI Agents with a bottoms-up, fine-grained per-agent model.

For each task that the business automates with a GUI Agent, there are 4 numbers that can be used to determine ROI:

• The time it takes for a human to complete the task (x)

• The salary/hr for the human (y)

• The time it takes for an agent to complete the same task (a)

• The cost of the agent/hr (b)

Then the ROI is very simply: (x*y) / (a*b)

With optimizations in the model and the GUIDE Harness, we find that the time for the agent to complete a task is often lower than the time it takes the human to complete the same task. Further, the cost of the tokens for the agent can range from $1.50/hr for the open source models to $15/hr for the frontier models.

GUIDE allows the business to choose the model on a per-agent basis. Therefore, if the GUI Agent works well to automate a task with a lower cost model, then the ROI for the task will be higher. Some very high value human tasks may need a more expensive frontier model.

Businesses can now compute the ROI of GUI Agents on a task-by-task basis.