Quartz 4

WR - Integrating AI Agents to Enhance Software Development Productivity and Quality

20 min read

_D_SWEL

From: Inquiry,General <inquiry@gartner.com>
Sent: Thursday, October 9, 2025 3:48 PM
To: Volmar Segundo <Volmar.Segundo@gartner.com>
Subject: Gartner - (Reference # 19002502) - WR - Integrating AI Agents to Enhance Software Development Productivity and Quality

CLIENT INQUIRY

Date: 09 Oct 2025Inquiry Reference: 19002502
Client Question: The client, André Nobre, Gerente de Sistemas at TJCE, seeks guidance on the effective integration of AI agents into existing software development processes. The goal is to leverage AI to drive improvements in team productivity and software quality while aligning with industry benchmarks and best practices. The client is exploring AI adoption to modernize workflows and optimize resource utilization.

Specific Questions:
What are the most effective use cases for AI agents in software development (e.g., code review, testing, documentation)?
Which industry benchmarks should be used to assess AI-driven productivity and quality improvements?
What change management approaches ensure successful AI integration with minimal disruption?
How can teams maintain software quality and security when leveraging AI agents?
What are the common challenges and pitfalls in AI adoption within software development, and how can they be addressed?

Additional Client Details: Client: André Nobre, Gerente de Sistemas, TJCE.
Focus: AI integration in software development processes.
Drivers: Productivity, quality, modernization, resource optimization.
Organization: Tribunal de Justiça do Ceará (TJCE).

Intended Outcome:

Understand proven strategies for integrating AI agents into software development lifecycles.
Identify benchmarks and KPIs for measuring the impact of AI on productivity and quality.
Learn best practices for AI implementation, including change management and governance.
Gain insights into potential risks and mitigation strategies associated with AI adoption.

Dear André,

Thank you for your inquiry on the “Integrating AI Agents to Enhance Software Development Productivity and Quality”.

AI agents are transforming software development by automating repetitive tasks (e.g., code generation, testing, documentation), improving code quality, and streamlining collaboration. Industry benchmarks such as DORA and SPACE frameworks, along with developer productivity insight platforms, help measure the impact of AI on productivity and quality. Successful integration requires phased adoption, robust change management, and ongoing upskilling, with a focus on prompt engineering and governance. Maintaining software quality and security involves rigorous code reviews, legal and security safeguards, and clear accountability for AI-generated code. Common challenges include unrealistic expectations, skill erosion, and security risks, which can be mitigated by balanced adoption, transparent culture, and continuous learning. AI augments but does not replace software engineers; instead, it drives demand for new hybrid roles and product-oriented mindsets. Leaders should prioritize value creation over cost reduction and apply AI across the full software development lifecycle to maximize business outcomes.

For further questions or wish to explore either section in greater detail, we recommend scheduling a consultation with our experts. It may happen that you do not have access to some of the resources based on your Gartner subscription, we request you contact your Gartner Account Executive (AE) to gain access to these resources. 

Effective Use Cases for AI Agents in Software Development

AI agents are increasingly being integrated into software development processes, offering a range of effective use cases that enhance productivity, streamline workflows, and improve overall software quality. Below are some of the most notable applications of AI agents in this field:

1. Automating Repetitive Tasks

  • Code Generation: AI agents can assist in generating code snippets based on natural language prompts, significantly speeding up the coding process. This includes generating boilerplate code, which allows developers to focus on more complex tasks.
  • Automated Testing: AI agents can run various tests automatically, including unit tests and integration tests, providing detailed reports on any issues found. This not only saves time but also enhances the reliability and quality of the software.

2. Enhancing Developer Experience

  • Streamlined Onboarding: AI agents can create and update comprehensive documentation automatically, helping new developers understand projects quickly.
  • Self-Service DevOps Workflows: AI agents can manage deployment pipelines, automatically handling rollbacks in case of failures and continuously monitoring application performance, thus reducing the manual workload for developers.

3. Improving Code Quality

  • Code Review and Refactoring: AI agents can assist in enforcing coding standards and best practices, such as adding error handling and maintaining documentation standards. This reduces the cognitive load on senior engineers and ensures junior developers benefit from established best practices.
  • Technical Debt Remediation: AI agents can systematically traverse codebases to identify and resolve deprecated API usage and assist with modernizing legacy code patterns, ensuring consistent application of updates across related components.

4. Facilitating Collaboration

  • Knowledge Sharing: AI agents can curate and share relevant code snippets, documentation, and best practices from a centralized repository, making it easier for team members to access the information they need.
  • Multiagent Systems: In complex workflows, multiple AI agents can collaborate to achieve shared goals, allowing for more efficient task execution and improved accuracy.

5. Supporting Decision-Making

  • User Feedback Analysis: AI agents can analyze user feedback from various channels to identify common issues and feature requests, suggesting iterative improvements to the product based on continuous monitoring of user interactions.
  • Planning and Discovery: AI agents can traverse and understand codebases, enabling faster project initiation and more informed technical decision-making by analyzing system dependencies and identifying potential integration points.

For more details, kindly access:

Value of GenAI Tools and AI Agents in Software Development

AI-augmented software development is poised to transform traditional development practices, with AI code assistants leading the change. A recent Gartner survey indicates that 18% of organizations have already deployed AI code assistants, with another 25% in the deployment stage, 20% in the piloting stage and 14% in the planning stage.

In the context of these expectations, software engineering leaders must determine ROI and build a business case as they scale their rollouts of AI code assistants. However, traditional ROI frameworks and vendor rhetoric steer engineering leaders toward metrics focused on cost reduction (and the implicit promise of reducing developer headcount). This narrow perspective fails to capture the full value of AI code assistants. How, then, can software engineering leaders build an effective value story that extends beyond traditional ROI metrics, manages stakeholder expectations and captures the full benefits of this technology investment?

The key lies in reframing the ROI conversation from cost reduction to value generation, thus capturing the full enterprise value story for AI code assistants (see Figure 1).

Figure 1: Value of AI Code Assistants

The limited view of the value of AI code assistants comprises faster coding, time savings and cost reduction. The additional value AI code assistants can provide includes enhancing developer experience, improving CX and improving developer retention, among others. This captures the full enterprise value story for AI code assistants.

Please review How to Communicate the Value of AI Code Assistants#. This research helps leaders articulate the breadth of benefits by reframing the conversation from cost reduction to value creation.

You can access our research note How to Calculate the Value and Cost of AI Agents. Leaders responsible for AI must evaluate the potential value and cost of new agentic AI use cases. This insight provides a framework for building and simulating total cost and evaluating key drivers of cost and value at scale for AI agent use cases.

You can also access our Market Guide for Developer Productivity Insight Platforms. Gartner defines developer productivity insight platforms as solutions that provide software engineering leaders with data-driven visibility into the engineering team’s use of time and resources, operational effectiveness, and progress on deliverables. This enables software engineering leaders and their teams to make smarter business decisions, resulting in higher developer productivity. Developer productivity insight platforms must be capable of ingesting and analyzing the abundant data created by common engineering tools and systems. They must provide rich, tailored, and role-specific user experiences to help leaders more easily identify constraints, spot important trends, and gain contextual insights.

Developer productivity insight platforms provide visibility of team activities with customizable dashboards and support metrics frameworks like DORA and SPACE. Software engineering leaders should use these platforms to drive improved productivity and value delivery.

Kindly access below resources on finding value of GenAI tools in software development:

GenAI Improving Developer Productivity

AI code assistants are being rapidly adopted across organizations, but their impact is limited to a small portion of the software development life cycle. Our research “Expand the Use of AI Tools Across Software Development to Maximize Productivity” advises software engineering leaders how to unlock broader productivity gains by expanding the usage of AI tools across teams and the SDLC.

Software engineering leaders operate in an environment where investment is increasing, and there is a focus on delivering software more rapidly. Only 36% of respondents in the Gartner Software Engineering Survey for 2025 reported being successful in reducing time to market. This brings us to a crucial consideration: What are the top practices for managing and improving software engineering productivity? Software engineering leaders can use our research A Practical Guide to Improving Developer Productivity to navigate these challenges and effectively measure and improve productivity to deliver greater business value.

You can also access our research note Boost Developer Productivity by Improving Adoption of AI Tools. Nearly all software engineering organizations offer AI tools to their developers, but adoption of and satisfaction with these tools is low. This research provides insights into areas of staff dissatisfaction with AI and offers proven practices to improve productivity.

GenAI promises productivity gains for software engineering teams, but leaders struggle to find the most valuable use cases and navigate training challenges and security risks. Software engineering leaders can use our research How Software Engineering Leaders Use GenAI to Boost Productivity to gain insights and lessons learned from peers implementing GenAI.

Uncovering Myths

Key Findings

  • Vendors’ claims of generative AI (GenAI) enhancing software development productivity by 55% are narrow and too specific, and they do not reflect the entire software engineering practice. Realistically, GenAI currently offers an overall productivity improvement of 10% to 15%, with a potential to go up to 30% in the next three years.
  • GenAI’s productivity impact varies significantly across different phases of the software development life cycle (SDLC). The development and testing phases experience the most substantial gains, with 20% to 40% and 30% to 40% improvements, respectively, while other phases like design and deployment have more modest benefits.
  • “Experience level of a software engineer” and “task/role complexity” are the two determining factors for optimal productivity from GenAI tools for software engineering.
  • The impact of GenAI is not uniform across all software engineering tasks. For instance, GenAI offers greater advantages in generating new code (with potential time savings of up to 50%) compared with code maintenance (with time savings of 10% to 30%). However, time savings and productivity gains must be translated into business value.

Figure 2: Staff Reporting More Than 10% Productivity Gains by Number of AI Use Cases in Software Development

A line chart showing the percentage of software engineering staff reporting more than 10% productivity gains by number of AI use cases. As the number of AI use cases increases, the percentage of staff reporting more than 10% productivity gains increases.

For more information, kindly access GenAI in Software Development: Uncovering Myths and Realities for Tech CEOs*.

For more stats, kindly access our Top Strategic Trends in Software Engineering for 2025.

GenAI Gains in Software Engineering

Please find attached snippets from our lates Gartner Software Engineering Survey 2025 to understand more on adoption of GenAI tools in software engineering. You can get on an inquiry call with our expert to walk you through the survey results.

AI is set to continue reshaping software engineering in 2025, but it is crucial to navigate the surrounding hype with caution. The opportunities with GenAI are vast, but there are risks to navigate. Leaders must avoid a narrow focus on coding and problem-solving and instead emphasize solid software engineering fundamentals. Organizations should restructure to support new team dynamics, where an increasing number of team members are likely to lack formal computer science training. GenAI should be applied across the full software development life cycle, not just in coding. Teams must be properly trained and supported to avoid tool misuse, which could lead to building the wrong product, accumulating technical debt, or making critical errors like data misuse. Software engineering leaders must strategically manage AI adoption to ensure that technological advancements translate into meaningful business outcomes. They must avoid the trap of chasing superficial productivity metrics — such as the number of lines of code or the number of tests — and instead apply AI tools across the entire software life cycle to ensure a balanced and systemic focus on business outcomes. This approach will harness AI’s potential while mitigating overreliance risks, aligning AI-driven efficiencies with strategic objectives, and balancing innovation with realistic expectations (see Figure 3).

Figure 3: Expected Versus Realizable Value From GenAI in Software Engineering

The chart illustrates the relationship between two trends: a steep upward trend in the expected value of GenAI, driven by the hype surrounding it, and a slower upward trend in the realizable value. The latter is lower due to the technical and organizational challenges involved in effectively integrating GenAI.

Kindly access Predicts 2025: Navigating the Rise of AI in Software Engineering for more information.

Generative AI Impacts (Benefits, Risks and Pitfalls) in the Software Delivery Life Cycle

Our research How Generative AI Impacts the Software Delivery Life Cycle* outlines how GenAI can assist in the development process and the risks and pitfalls you must manage when using it, as illustrated in Figure 4. 

Figure 4: Balancing GenAI Benefits and Risks

GenAI benefit areas include code generation, debugging and modernization, as well as AI-augmented testing and artifact building. However, there are some risks and pitfalls. These include confirmation bias, loss of skills, reduced code quality and lack of adoption.

This research covers three areas:

Kindly access the above linked research for further details.

Maintaining Quality, Managing Risk, Protecting IP for AI-Generated Code

Kindly access the resources below:

Developing Prompt Engineering

Figure 5 illustrates the main components of developing the skill of prompt engineering. Software engineers need to be able to take business needs such as user stories and translate them into prompts that will derive the required LLM behaviors.

Figure 5: Skills for Prompt Engineering

The three main skills needed for prompt engineering are business (high-involvement), data (moderate-involvement) and AI (low-involvement) skills. Combining business acumen and sample data can allow nontechnical business users who learned those skills to perform better.

Jump-start your prompting skills with technical LLM collaboration, and experiment with text-to-application platforms such as:

These tools demonstrate the potential of using AI to build software, but consider them as proving grounds for refining your prompting skills as you seek to improve your use of LLMs to optimize code development. Leverage “prompt rewrite” features to better understand how AI can optimize your initial prompts. Do not assume AI-generated code, libraries and architectures are perfect; be skeptical of the design styles and systems assumed by AI.

Build your own milestone-based curriculum over the next 12 months to learn GenAI for software engineering. Set a goal of becoming a master at software-engineering-related LLM prompting. Include a curated software engineering and GenAI course. Set realistic calendar reminders to complete the material, and document the techniques from the course and from your meaningful code assistant interactions and LLM experiments.

Get started with software-development-centric GenAI courses, such as:

Prompt engineering skills are only growing in demand as more applications incorporate GenAI. For more details on how to develop prompt engineering skills, refer to the following Gartner research:

For more information, kindly access Essential Skills for Building Generative AI Applications*.

Impact of AI on Software Engineering Roles and Upskilling

Why AI Will Not Replace Software Engineers?

According to 2025 World Economic Forum data, software developer roles are projected to grow by 57% from 2025 to 2030, with AI cited as a key driver. You can refer to Future of Software Engineering 2030: The Impact of AI to gain insights upon Six Gartner Positions on the Future of Software Engineering in detail.

Rumors of the demise of software engineers are greatly exaggerated, despite generative AI’s early promise. By understanding AI’s impact in this domain, executive leaders can develop a more grounded perspective on how it will transform work across their organization.

AI will affect software engineering in different ways over time (see Figure 6):

  • In the short term, AI tools augment specific software development tasks. Productivity gains are modest and unevenly distributed.
  • In the medium term, AI tools will become more “agentic,” meaning they will be able to break down complex problems into discrete subtasks that can be fully automated. To guide AI agents, software engineers will need to develop new skills in prompt engineering and retrieval-augmented generation (RAG).
  • In the long term, the efficiency gains and new capabilities enabled by AI will increase demand for complex, innovative software. That will drive demand for a new breed of highly skilled human AI engineers.

 Figure 6: AI Disruption Spectrum

AI disruption spectrum anticipates the degree of work disruption of AI investments. The spectrum has three stages and AI has transformative impacts: within boundaries (augments existing work patterns and tasks), pushing boundaries (transforms work patterns) and breaking boundaries (transforms work patterns and its types).

In the long term, AI will introduce efficiency gains and enable new capabilities that fuel an ever-growing demand for complex, innovative software. Although AI will enable software engineers to be more productive, organizations will actually need even more skilled software engineers to meet the rapidly increasing demand for AI-empowered software. This phenomenon is known as the Jevons Paradox. As resource efficiency improves, it stimulates demand and expands the scope of resource utilization instead of reducing overall usage.

For example, the introduction of fuel-efficient cars did not reduce energy usage. Instead, it reduced fuel costs for consumers; and because fuel was cheaper, people started driving more. Fuel-efficient cars actually led to an increase in energy consumption. Although it seems counterintuitive, the increased productivity of AI-native software engineering teams will drive demand for AI-empowered applications and, thus, increase demand for software engineers.

Figure 7: Jevons Paradox: Improved Efficiency Leads to Greater Demand

A diagram of a diagram AI-generated content may be incorrect.

Kindly access “Why AI Will Not Replace Software Engineers, and What That Means for the C-Suite” for more information. Additionally, you may also leverage AI Will Not Replace Software Engineers (and May, in Fact, Require More) for more insights.

Insights on AI’s Impact on Roles, Skills, Engineering Culture

  • AI is shifting how teams deliver value: With GenAI taking over tasks like boilerplate coding and documentation, software engineering teams must shift from output-focused roles to a product-oriented mindset — prioritizing outcomes, iteration and customer value over raw code volume.

  • New hybrid roles are emerging: Teams are evolving beyond full-stack and DevOps into roles like,

  • GenAI app integrator

  • AI engineer

  • The developer experience is under pressure: Engineers face rising cognitive load, role ambiguity, and tooling sprawl. Without deliberate orchestration, productivity gains promised by AI can be eroded by fatigue and fragmentation.

  • Organizational culture must support constant reinvention: High-performing teams in the AI era will be those that normalize experimentation, treat failure as learning and support transparent, cross-functional collaboration.

Please leverage AI Playbook for Software Engineering Leaders for more insights. You can refer to Slides 10-12 in this playbook for gaining more insights upon recommended actions for managing software engineering teams to navigate AI impact upon software engineering roles.

AI’s General Impact on Software Development and Developer Productivity

Core Changes: The industry is shifting from fixed, sequential LLM workflows to more dynamic, multi-agent supported systems. An AI agent is an autonomous entity capable of perceiving its environment, making decisions, using tools and acting to achieve goals. This evolution necessitates robust orchestration frameworks to manage the intricate workflows, state and interactions of these AI agents. In the SDLC this usually means bounding their area of impact and carefully guard railing their actions until they reach a consistent, robust and cost-effective level of operation.

Impact on the Classic Software Engineer Workflow: AI-powered tools like code assistants, improve the developer experience (DevEx) and productivity. Engineers spend less time on repetitive tasks such as boilerplate code generation, routine unit tests, writing documentation and light refactoring. Particularly good for greenfield projects. This allows developers to focus more on core application logic and complex problem-solving. While vendor claims suggest 40-55% gains, real world productivity increases are a more modest but still significant 8-20%, which compounds over time, mosty from timesaving. AI also augments various stages of the Software Development Life Cycle (SDLC), particularly coding and testing, with emerging applications helping with requirements gathering, design/user stories, code review and post-deployment monitoring. They tend to contribute to reduced code merge times, improved deployment frequency and improved velocity.

Important Considerations: Developers must remain fully accountable for all code, whether human written or AI-generated, necessitating rigorous code reviews to catch AI-introduced errors, security vulnerabilities, or “hallucinations.” The effort required for prompt engineering and validating AI generated code can offset some time savings or need to refactor. There is a risk of skills diminishing in areas automated by AI, making it crucial for developers to evolve towards a system or product engineering mindset. Security and intellectual property concerns arise from sending code and data to model providers, requiring legal and security approvals and potentially air-gapped or Virtual Private Cloud (VPC) AI implementations for sensitive data. Effectively managing unrealistic expectations and encouraging phased adoption are also key to balance gains with risks.

Recommendations to Navigate This Impact on Roles

Below are some recommended reads to help you navigate AI impact upon software engineering roles:

  • AI is expected to impact every facet of how software engineering organizations work, what they build and how they build it. The pace and unpredictability of AI-driven change leaves many software engineering leaders and teams struggling to keep up. This note provides critical insight on how to navigate AI-driven changes successfully based on Gartner proprietary survey data. Please leverage How Engineering Leaders Expect AI to Disrupt Software Development for more insights.

  • Organizations that fail to evolve their software developers from coders to digital conductors who orchestrate code development will be unable to compete in the AI era. Software engineering leaders must invest in five foundational practices to set up their teams for AI-native engineering success. Please leverage Software Engineering Foundations for the AI-Native Era for more insights.

  • As AI tools reshape development and automate entry-level tasks, software engineering leaders face a critical skills-experience paradox. Success requires balancing AI tool adoption with foundational skills development while creating new pathways for practical experience. Please leverage How to Upskill Software Engineering Teams in the Age of AI for more insights.

  • Software engineering teams are under pressure to include AI features in custom applications but often lack the required AI skills. Software engineering leaders can use the data and analysis in this research to learn common approaches to upskilling and hiring for AI expertise. Kindly access How to Build AI Skills in Software Engineering Teams for more information.

Additional Note(s) to Refer:

Note: If images are not displayed on your device, please try viewing them on a laptop. Alternatively, consult the associated research note for the necessary information.

*_ Note_**: If you do not have access to any of the research notes mentioned in the response that you wish to read in detail, we recommend you contact your Gartner representative to gain access.

# Note: This research note is archived and some of the insights may not be relevant as of today

Graph View

Backlinks