AI Tools in Software Development: Developer Perceptions and Usage Patterns

Mark Looi

arxiv: 2601.21305 · v4 · submitted 2026-01-29 · 💻 cs.SE

AI Tools in Software Development: Developer Perceptions and Usage Patterns

Mark Looi This is my paper

Pith reviewed 2026-05-16 10:06 UTC · model grok-4.3

classification 💻 cs.SE

keywords AI toolssoftware developmentproductivitycode qualitydeveloper perceptionssurvey analysistechnology adoption

0 comments

The pith

Higher AI tool usage links to higher perceived productivity and code quality among developers.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines survey responses from 147 professional developers to map how often and how widely they use generative AI tools in coding and testing. It finds that more frequent and broader use corresponds to reports of both higher productivity and higher code quality, with no perceived trade-off between the two. Adoption intentions rise with current usage, ease of integration, and organizational policies, while security concerns exert a mild drag. Usage in testing lags behind coding, and the responses cluster into three groups that differ in their optimism and context.

Core claim

Analysis of the 147 responses shows that frequency and breadth of AI tool use are positively associated with perceived productivity gains and perceived code quality, and that these two outcomes co-occur rather than trade off. Adoption intent correlates with existing usage patterns and integration ease, while security concerns show only modest negative association. Coding tasks see heavier AI use than testing tasks, and the data separate into three developer segments—Enthusiasts, Pragmatists, and Cautious—that differ in usage breadth, outcome perceptions, and future adoption plans, patterns that align with organizational AI policies.

What carries the argument

Survey-based associations between AI tool usage frequency and breadth on one side and self-reported productivity plus code quality on the other, with k-means clustering producing three usage-perception segments.

If this is right

Developers who integrate AI tools more deeply into daily work report simultaneous gains in speed and quality.
Lower AI adoption in testing suggests targeted tool improvements could raise productivity in that area.
Ease of integration and existing usage patterns predict stronger future adoption intent.
Organizational AI policies correlate with differences across the three developer segments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If perceptions track reality, organizations could accelerate adoption by reducing integration friction and clarifying policies.
The gap between coding and testing usage points to a possible opportunity for tools that better support verification workflows.
The three segments imply that uniform rollout strategies may underperform compared with tailored support for cautious users.

Load-bearing premise

Self-reported survey answers from the 147 developers accurately reflect their actual productivity, code quality, and usage habits without meaningful recall or social-desirability bias.

What would settle it

A controlled study that tracks objective metrics such as task completion time, defect rates, and code review scores for the same developers working on matched tasks with and without AI tools.

read the original abstract

The use of Generative AI (GenAI) tools in software development has raised questions about their impact on productivity, code quality, and developer practices. Prior research presents mixed findings, with objective analyses identifying potential declines in code quality, while survey-based studies report perceived improvements in productivity and minimal quality trade-offs. This study presents an empirical analysis of survey data from 147 professional developers, examining associations between AI tool usage, perceived productivity, perceived code quality, and adoption intent. The results indicate that higher frequency and broader use of AI tools are associated with higher perceived productivity and perceived code quality. In contrast to concerns about a trade-off between speed and quality, developers report that these outcomes co-occur. Adoption intent is positively associated with current usage patterns and ease of integration, while security concerns show a modest negative association. AI tool usage in testing is less extensive than in coding, and perceived productivity gains are correspondingly lower. Clustering analysis identifies three developer segments-Enthusiasts, Pragmatists, and Cautious-which differ in usage breadth, perceived outcomes, and adoption intent. These patterns are consistent with diffusion-like models of technology adoption and suggest a reinforcing relationship between usage, perceived outcomes, and intended future use. These segments are also associated with differences in reported organizational context, including the presence of AI-related policies. Overall, the findings suggest that perceptions of productivity and quality, along with usage experience and integration factors, are associated with adoption behavior, highlighting the importance of considering both perceived and objective outcomes in evaluating AI-assisted software development.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This survey of 147 developers links higher AI tool use to better perceived productivity and quality with no reported trade-off, but the single-source self-report design leaves the associations open to bias.

read the letter

The paper's core finding is that among these 147 professional developers, more frequent and broader AI tool use lines up with higher self-reported productivity and code quality, and respondents see those two outcomes moving together rather than trading off. The clustering into Enthusiasts, Pragmatists, and Cautious groups, plus the links to organizational policies and adoption intent, gives the results some structure that prior mixed studies lacked. The split between coding and testing use is also a concrete detail worth noting. The work is straightforward empirical survey analysis with no circular derivations or invented parameters, and the associations are drawn directly from the collected responses. That keeps the claims grounded in the data they actually have. The main limitation is that everything—usage, productivity, quality, and intent—comes from the same respondents at the same moment. Common-method bias or social-desirability effects could easily produce the positive pattern and the lack of trade-off, and the paper offers no objective metrics, multi-source checks, or follow-up data to test that. The reinforcing-relationship interpretation therefore stays suggestive rather than demonstrated. This is useful for teams trying to gauge how developers are actually reacting to AI tools right now, or for researchers planning similar perception studies. A serious referee could push on the methods for response-rate details, statistical controls, and bias checks, which would strengthen it. I would send it to review rather than desk-reject.

Referee Report

2 major / 2 minor

Summary. The paper reports results from a survey of 147 professional developers on Generative AI tool usage in software development. It finds that higher frequency and broader use of AI tools are associated with higher perceived productivity and perceived code quality, with these outcomes co-occurring rather than trading off. Adoption intent correlates positively with usage patterns and integration ease but negatively with security concerns. Clustering identifies three segments (Enthusiasts, Pragmatists, Cautious) differing in usage, perceptions, and intent, consistent with diffusion models and linked to organizational AI policies. Usage in testing lags coding, with correspondingly lower perceived gains.

Significance. If the associations prove robust, the work adds to software engineering literature by mapping developer segments and reinforcing relationships between usage, perceptions, and adoption intent. It highlights practical factors like integration ease and policy presence that could guide tool deployment and organizational strategies, while aligning survey patterns with established technology diffusion frameworks.

major comments (2)

The central claims of positive associations between AI usage frequency/breadth and perceived productivity/quality (with no trade-off) rest entirely on cross-sectional self-reports from a single instrument. This setup is vulnerable to common-method bias and social desirability, which could artifactually produce the reported correlations and segment distinctions. No objective metrics, multi-rater validation, or longitudinal data are provided to establish construct validity. (Abstract and Results sections describing the associations and clustering)
The manuscript does not report response rate, sampling frame details, or any statistical controls for confounders (e.g., experience level, team size, or organization type). These omissions limit interpretation of the generalizability of the Enthusiasts/Pragmatists/Cautious segments and the adoption-intent findings. (Methods section)

minor comments (2)

Clarify the clustering procedure (algorithm, number of clusters determination, validation metrics such as silhouette score) to support reproducibility of the three-segment solution.
Add explicit discussion of study limitations, including reliance on perceptions and absence of objective code-quality measures, to balance the interpretation of the reinforcing-relationship claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and have revised the manuscript to improve transparency and acknowledge limitations where feasible.

read point-by-point responses

Referee: The central claims of positive associations between AI usage frequency/breadth and perceived productivity/quality (with no trade-off) rest entirely on cross-sectional self-reports from a single instrument. This setup is vulnerable to common-method bias and social desirability, which could artifactually produce the reported correlations and segment distinctions. No objective metrics, multi-rater validation, or longitudinal data are provided to establish construct validity. (Abstract and Results sections describing the associations and clustering)

Authors: We acknowledge that the study is based on cross-sectional self-reported data and is thus susceptible to common-method bias and social desirability effects. Perceptual measures of productivity and quality are standard in software engineering research on tool adoption, as objective metrics are difficult to obtain at scale without experimental controls. The contribution lies in identifying patterns in developer perceptions and their links to adoption intent, consistent with prior survey-based work. We have expanded the Limitations section to discuss these biases explicitly, added stronger caveats to the abstract and results, and clarified that no objective or longitudinal data were collected given the survey design focused on current perceptions. revision: partial
Referee: The manuscript does not report response rate, sampling frame details, or any statistical controls for confounders (e.g., experience level, team size, or organization type). These omissions limit interpretation of the generalizability of the Enthusiasts/Pragmatists/Cautious segments and the adoption-intent findings. (Methods section)

Authors: We have revised the Methods section to report the response rate (28% from the recruitment channels), provide full sampling frame details (professional developers recruited via targeted LinkedIn groups and developer forums with screening for current employment in software roles), and include additional regression controls for confounders such as years of experience, team size, and organization type. These changes enhance interpretability of the segments and adoption findings. revision: yes

Circularity Check

0 steps flagged

No significant circularity in empirical survey analysis

full rationale

The paper conducts an empirical analysis of cross-sectional survey responses from 147 developers, reporting statistical associations between self-reported AI tool usage frequency/breadth and perceived productivity/quality outcomes, along with clustering into usage-based segments. No mathematical derivations, equations, fitted parameters, or self-citations are present that reduce any central claim to its own inputs by construction. All reported patterns (positive associations, co-occurrence of speed and quality perceptions, segment differences) are computed directly from the collected data without self-definitional loops, renamed predictions, or load-bearing uniqueness theorems imported from prior author work. The analysis is therefore self-contained against external benchmarks and receives a score of 0.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Central claims rest on standard survey methodology assumptions including honest self-reporting and representative sampling of professional developers; no free parameters or invented entities introduced.

axioms (2)

domain assumption Self-reported perceptions of productivity and code quality accurately reflect actual outcomes
Invoked throughout results section when interpreting associations from survey responses
domain assumption Survey sample of 147 developers is sufficient to identify stable usage patterns and segments
Basis for clustering analysis and generalization to broader developer population

pith-pipeline@v0.9.0 · 5566 in / 1252 out tokens · 24825 ms · 2026-05-16T10:06:01.876871+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

higher frequency and broader use of AI tools are associated with higher perceived productivity and perceived code quality... clustering analysis identifies three developer segments—Enthusiasts, Pragmatists, and Cautious

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

[4]

PQI PCA + K-Means Clustering (k=3) Table 5: Futures Hypotheses In sum, our hypotheses will journey along the path:

work page
[5]

HP: Connect usage to accuracy, breadth, frequency, and productivity (PP)

work page
[6]

HQ: Assess if that tools breadth correlates with better perceived quality (PQI)

work page
[7]

HA: Assess if that PP and PQ are associated with the intent to use more AI

work page
[8]

Often” or “Always

HF: Assess if how these steps and our distinct developer archetypes illustrate an innovation diffusion process for AI in software engineering. 5.5 Data Collection This study is founded on a comprehensive survey dataset collected from over 170 software or software adjacent professionals. The data were further culled to only include those actively engaged i...

work page
[9]

Intent to Increase Usage Index

work page
[10]

Strategic Outlook Index

work page
[11]

AI Coding Tool Index (Usage Breadth)

work page
[12]

E=56 P=53 C=38 N/A N/A Table 17: Futures Hypotheses Findings Hypothesis HF7 was supported, revealing three distinct profiles based on survey responses

PQI 3 Archetypes. E=56 P=53 C=38 N/A N/A Table 17: Futures Hypotheses Findings Hypothesis HF7 was supported, revealing three distinct profiles based on survey responses. The clustering analysis used our four aggregate indices, which together combine 32 variables: Intent to Increase Usage Index, Strategic Outlook Index, AI Coding Tool Index (a measure of u...

work page 1989
[13]

This indicates that optimism about the future is grounded in the hands-on experience of building the next generation of AI-powered applications

The single strongest associate of a positive strategic outlook is engagement in AI-native development (HF3). This indicates that optimism about the future is grounded in the hands-on experience of building the next generation of AI-powered applications. Developers in the Age of AI: Adoption, Policy, and Diffusion of AI Software Engineering Tools 21

work page
[14]

Archetypes: In addressing RQ5, our clustering analysis identified three distinct developer archetypes: the Cautious, Pragmatists, and Enthusiasts, Table 18

The positive correlation between an intent to increase current AI usage and a positive strategic outlook (HF6) reinforces the virtuous adoption cycle—those who believe in AI’s future importance are the same ones deeply committed to increasing its use today. Archetypes: In addressing RQ5, our clustering analysis identified three distinct developer archetyp...

work page
[15]

They validate the success of the current generation of tools

Enthusiasts: Enthusiasts represent the vanguard, scoring highest on usage, outlook, optimism, and satisfaction with current PQ. They validate the success of the current generation of tools. They lead in almost all positive metrics. They are differentiated from the other archetypes by high AI coding Policy support (59%)

work page
[16]

They have the mid-level mean score for the perceived Impact of AI on Code Quality

Cautious: This archetype is defined by below-average usage and adoption metrics. They have the mid-level mean score for the perceived Impact of AI on Code Quality. Juxtaposed with their lower usage, this suggests they are doubtful that AI improves PQ (or they believe their current, non-AI code is adequate, leaving little room for AI to improve). They also...

work page
[17]

They are the most skeptical of current quality (PQI 3.40)

Pragmatists: This group is the most critical group for the future, matching Enthusiasts’ high optimism and intent though their adoption is moderate. They are the most skeptical of current quality (PQI 3.40). They are average in terms of tool usage Breadth. Pragmatists report the highest Intent to increase usage (Intent 4.51). They report limited organizat...

work page 2003
[18]

Thus, managers should treat AI tools as a performance accelerator and measure both productivity and quality to reinforce the feedback loop

Improve PP through Accuracy and Integration Developers in the Age of AI: Adoption, Policy, and Diffusion of AI Software Engineering Tools 23 Question the Quality Paradox: The study finds that PP-Code and PP-Test are positively correlated with PQI (HP1, HP7) (He, et al., 2025). Thus, managers should treat AI tools as a performance accelerator and measure b...

work page 2025
[19]

So, organizations should invest in training in these technologies

Realign Skills and Architecture for the AI-Native Era Prioritize Orchestration and Agentic: The future directions are influenced by beliefs about the importance of Agentic Architecture (Table 15) and Orchestration Workflow Design (HF1). So, organizations should invest in training in these technologies

work page
[20]

Notwithstanding their prudent concern about AI, Enthusiasts are vanguards of the AI transformation

Archetypes The presence of Archetypes (HF7) enables an engagement strategy for management to profile their teams to identify Enthusiasts, Pragmatists, the Cautious and to assist in the diffusion of new AI technologies: • Enthusiasts: Provide early access to Agentic and Orchestration systems to leverage their momentum. Notwithstanding their prudent concern...

work page
[21]

Early: If the organization lacks AI policies and adoption is low, an organization is likely in an early phase where Enthusiasts have not yet demonstrated success

Leverage the Diffusion Process Organizations progress through a diffusion process in AI adoption; they should recognize what works at each stage. Early: If the organization lacks AI policies and adoption is low, an organization is likely in an early phase where Enthusiasts have not yet demonstrated success. The answer is not to create Policy hoping it wil...

work page
[22]

Developers must strive for consistent, daily usage to meet or exceed the reported PP-Code gains (4 hours saved per week)

Habitual Use for Personal Productivity High frequency of AI tool use is the strongest single PP factor in both coding and testing (HP3, HP6). Developers must strive for consistent, daily usage to meet or exceed the reported PP-Code gains (4 hours saved per week). Breadth of application is the most consistent statistical factor for both PP-Code (HP2) and P...

work page
[23]

Still, the future belongs to Orchestration Workflow Design (HF1), RAG, and Agentic Architectures (Appendix Figure 20)

Master Prompt Engineering, Orchestration and Data-Grounding Prompt Engineering’s importance to the Futures Index (HF2) confirms that interaction with GenAI is a foundational design skill. Still, the future belongs to Orchestration Workflow Design (HF1), RAG, and Agentic Architectures (Appendix Figure 20). Developers should dedicate training time to these ...

work page
[24]

shadow work

Know your Archetype Developers can gain career clarity by understanding their usage and belief profiles (HF7) and recognize that these are malleable. 7.6 Further Research The current findings provide a robust snapshot of perceived AI tools adoption, productivity, and outlook in software development. To build upon these results, future research should try ...

work page
[25]

Adoption for testing is narrower, suggesting a comparative lack of maturity in that ecosystem

The Virtuous Adoption Cycle: The study confirms that the efficacy of AI tools correlates with adoption, engendering a self-reinforcing loop of usage, trust, and gains: • AI tools adoption is ubiquitous for coding, frequently and broadly used. Adoption for testing is narrower, suggesting a comparative lack of maturity in that ecosystem. • Developers realiz...

work page
[26]

Future Readiness: The developer community is highly optimistic and grounded in a belief that complex, multi-model systems will define the future of software development: • Developers view Orchestration or the ability to design, sequence, and manage multiple AI agents and data flows as the most critical future skill, followed by data management and Prompt ...

work page
[27]

Perceived usefulness, perceived ease of use, and user acceptance of information technology

Organizational Diffusion Process: AI adoption in professional software development unfolds as an organizational maturity and diffusion of innovations process. • Early adopters (Enthusiasts) push ahead with tools and accumulate success. Organizations eventually respond by formalizing policies because Policy indicates maturity. • These policies then functio...

work page doi:10.1145/3582083 1964

[1] [4]

PQI PCA + K-Means Clustering (k=3) Table 5: Futures Hypotheses In sum, our hypotheses will journey along the path:

work page

[2] [5]

HP: Connect usage to accuracy, breadth, frequency, and productivity (PP)

work page

[3] [6]

HQ: Assess if that tools breadth correlates with better perceived quality (PQI)

work page

[4] [7]

HA: Assess if that PP and PQ are associated with the intent to use more AI

work page

[5] [8]

Often” or “Always

HF: Assess if how these steps and our distinct developer archetypes illustrate an innovation diffusion process for AI in software engineering. 5.5 Data Collection This study is founded on a comprehensive survey dataset collected from over 170 software or software adjacent professionals. The data were further culled to only include those actively engaged i...

work page

[6] [9]

Intent to Increase Usage Index

work page

[7] [10]

Strategic Outlook Index

work page

[8] [11]

AI Coding Tool Index (Usage Breadth)

work page

[9] [12]

E=56 P=53 C=38 N/A N/A Table 17: Futures Hypotheses Findings Hypothesis HF7 was supported, revealing three distinct profiles based on survey responses

PQI 3 Archetypes. E=56 P=53 C=38 N/A N/A Table 17: Futures Hypotheses Findings Hypothesis HF7 was supported, revealing three distinct profiles based on survey responses. The clustering analysis used our four aggregate indices, which together combine 32 variables: Intent to Increase Usage Index, Strategic Outlook Index, AI Coding Tool Index (a measure of u...

work page 1989

[10] [13]

This indicates that optimism about the future is grounded in the hands-on experience of building the next generation of AI-powered applications

The single strongest associate of a positive strategic outlook is engagement in AI-native development (HF3). This indicates that optimism about the future is grounded in the hands-on experience of building the next generation of AI-powered applications. Developers in the Age of AI: Adoption, Policy, and Diffusion of AI Software Engineering Tools 21

work page

[11] [14]

Archetypes: In addressing RQ5, our clustering analysis identified three distinct developer archetypes: the Cautious, Pragmatists, and Enthusiasts, Table 18

The positive correlation between an intent to increase current AI usage and a positive strategic outlook (HF6) reinforces the virtuous adoption cycle—those who believe in AI’s future importance are the same ones deeply committed to increasing its use today. Archetypes: In addressing RQ5, our clustering analysis identified three distinct developer archetyp...

work page

[12] [15]

They validate the success of the current generation of tools

Enthusiasts: Enthusiasts represent the vanguard, scoring highest on usage, outlook, optimism, and satisfaction with current PQ. They validate the success of the current generation of tools. They lead in almost all positive metrics. They are differentiated from the other archetypes by high AI coding Policy support (59%)

work page

[13] [16]

They have the mid-level mean score for the perceived Impact of AI on Code Quality

Cautious: This archetype is defined by below-average usage and adoption metrics. They have the mid-level mean score for the perceived Impact of AI on Code Quality. Juxtaposed with their lower usage, this suggests they are doubtful that AI improves PQ (or they believe their current, non-AI code is adequate, leaving little room for AI to improve). They also...

work page

[14] [17]

They are the most skeptical of current quality (PQI 3.40)

Pragmatists: This group is the most critical group for the future, matching Enthusiasts’ high optimism and intent though their adoption is moderate. They are the most skeptical of current quality (PQI 3.40). They are average in terms of tool usage Breadth. Pragmatists report the highest Intent to increase usage (Intent 4.51). They report limited organizat...

work page 2003

[15] [18]

Thus, managers should treat AI tools as a performance accelerator and measure both productivity and quality to reinforce the feedback loop

Improve PP through Accuracy and Integration Developers in the Age of AI: Adoption, Policy, and Diffusion of AI Software Engineering Tools 23 Question the Quality Paradox: The study finds that PP-Code and PP-Test are positively correlated with PQI (HP1, HP7) (He, et al., 2025). Thus, managers should treat AI tools as a performance accelerator and measure b...

work page 2025

[16] [19]

So, organizations should invest in training in these technologies

Realign Skills and Architecture for the AI-Native Era Prioritize Orchestration and Agentic: The future directions are influenced by beliefs about the importance of Agentic Architecture (Table 15) and Orchestration Workflow Design (HF1). So, organizations should invest in training in these technologies

work page

[17] [20]

Notwithstanding their prudent concern about AI, Enthusiasts are vanguards of the AI transformation

Archetypes The presence of Archetypes (HF7) enables an engagement strategy for management to profile their teams to identify Enthusiasts, Pragmatists, the Cautious and to assist in the diffusion of new AI technologies: • Enthusiasts: Provide early access to Agentic and Orchestration systems to leverage their momentum. Notwithstanding their prudent concern...

work page

[18] [21]

Early: If the organization lacks AI policies and adoption is low, an organization is likely in an early phase where Enthusiasts have not yet demonstrated success

Leverage the Diffusion Process Organizations progress through a diffusion process in AI adoption; they should recognize what works at each stage. Early: If the organization lacks AI policies and adoption is low, an organization is likely in an early phase where Enthusiasts have not yet demonstrated success. The answer is not to create Policy hoping it wil...

work page

[19] [22]

Developers must strive for consistent, daily usage to meet or exceed the reported PP-Code gains (4 hours saved per week)

Habitual Use for Personal Productivity High frequency of AI tool use is the strongest single PP factor in both coding and testing (HP3, HP6). Developers must strive for consistent, daily usage to meet or exceed the reported PP-Code gains (4 hours saved per week). Breadth of application is the most consistent statistical factor for both PP-Code (HP2) and P...

work page

[20] [23]

Still, the future belongs to Orchestration Workflow Design (HF1), RAG, and Agentic Architectures (Appendix Figure 20)

Master Prompt Engineering, Orchestration and Data-Grounding Prompt Engineering’s importance to the Futures Index (HF2) confirms that interaction with GenAI is a foundational design skill. Still, the future belongs to Orchestration Workflow Design (HF1), RAG, and Agentic Architectures (Appendix Figure 20). Developers should dedicate training time to these ...

work page

[21] [24]

shadow work

Know your Archetype Developers can gain career clarity by understanding their usage and belief profiles (HF7) and recognize that these are malleable. 7.6 Further Research The current findings provide a robust snapshot of perceived AI tools adoption, productivity, and outlook in software development. To build upon these results, future research should try ...

work page

[22] [25]

Adoption for testing is narrower, suggesting a comparative lack of maturity in that ecosystem

The Virtuous Adoption Cycle: The study confirms that the efficacy of AI tools correlates with adoption, engendering a self-reinforcing loop of usage, trust, and gains: • AI tools adoption is ubiquitous for coding, frequently and broadly used. Adoption for testing is narrower, suggesting a comparative lack of maturity in that ecosystem. • Developers realiz...

work page

[23] [26]

Future Readiness: The developer community is highly optimistic and grounded in a belief that complex, multi-model systems will define the future of software development: • Developers view Orchestration or the ability to design, sequence, and manage multiple AI agents and data flows as the most critical future skill, followed by data management and Prompt ...

work page

[24] [27]

Perceived usefulness, perceived ease of use, and user acceptance of information technology

Organizational Diffusion Process: AI adoption in professional software development unfolds as an organizational maturity and diffusion of innovations process. • Early adopters (Enthusiasts) push ahead with tools and accumulate success. Organizations eventually respond by formalizing policies because Policy indicates maturity. • These policies then functio...

work page doi:10.1145/3582083 1964