AI Assistance Usage Delphi Form


This Delphi form is for collecting data to quantify the impact of AI assistance on software development cost. It will be used to derive Effort Multipliers (EMs) for the proposed factor AI Assistance Usage similar to the impact for Applications Experience below from the COCOMO II model [1]. application experience.png

The effort multipliers for each rating represent the relative effort to Nominal. For example, the EM for a Very Low rating of Applications Experience is 1.22 indicating a 22% increase in effort from Nominal. The Low rating EM is 1.1 for a 10% increase in effort from Nominal. The Nominal rating is always 1.0 by definition for a typical project. The High rating EM is 0.88, or a 12% decrease in effort from Nominal. Very High EM is 0.81 for a decrease of 19% effort from Nominal. The overall Effort Multiplier Ratio (EMR) for Applications Experience is the ratio of the highest to lowest multipliers, or 1.22/.81 = 1.5.

Select only one of the three methods to submit your data. Method 1 asks for your effort multiplier estimates for each of the ratings. Method 2 asks for an estimate of the Effort Multiplier Ratio. Method 3 requires actual effort data using AI assistance along with an effort estimate (or actual) of developing the same without AI.

We are also interested in feedback on the rating scheme. Provide additional information if the Nominal rating for a typical project should be redefined based on your experience, or any of the others.

Thank you for your interest and support. Optionally provide your name for followup on this survey and research. Send questions to

Boehm Center for Systems and Software Engineering

Data Form

The table below defines the proposed AI Assistance Usage factor ratings for which we desire effort multipliers.

AI Assistance Usage Ratings

Very Low Low Nominal High Very High
• Minimal to no AI assistance.
• Development relies primarily on traditional methods and tools.
• AI tools may be present but are rarely, if ever, consulted.
• Occasional AI consultation, typically for clarification or basic information retrieval.
• AI tools are not deeply integrated into the development workflow.
• Regular use of AI tools for various tasks like code help, design insights, or testing assistance.
• AI tools are a recognized part of the toolkit but aren't central to development.
• Frequent and strategic use of AI assistance.
• AI tools play a central role in multiple phases of development, from design to code review.
• AI tools are deeply ingrained in most development phases. They are crucial for decision making, problem solving, and automating specific tasks.
• The development process is designed around maximizing AI tool benefits.
Select only one method:
Method 1: Estimated Effort Multipliers
What are your estimated effort multipliers for each rating relative to Nominal set to 1.0?
Method 2: Estimated Effort Multiplier Ratio (EMR)
Provide your estimated effort ratio from Very Low to Very High. E.g., the overall EMR for Applications Experience is the ratio of 1.22/.81 = 1.5 for the multiplicative range. If you provide an effort ratio between two other settings then explain in Rationale and Supporting Information.
Method 3: Effort Data from Project, Experiment or Case Study
Select the AI Assistance Usage rating applicable to the development:
Actual effort with AI assistance: Person-hours
Estimated effort without AI assistance: Person-hours

Please add a note if your effort without AI assistance is actual instead of estimated.

Rationale and Supporting Information:

[1] B. Boehm, C. Abts, W. Brown, S. Chulani, B. Clark, E. Horowitz, R. Madachy, D. Reifer, and B. Steece. Software Cost Estimation with COCOMO II. Prentice-Hall, 2000.