This post summarizes our Use Case Spotlight session on 22nd January, 2025, featuring Generative AI use cases. Our first session in the Generative AI series covered its application in software development, with respect to the Software Development Life Cycle (SDLC) phases.
More...
Top Trends in AI for Software Development
Artificial intelligence is transforming software development. There has been a stupendous rise in the AI coding assistants that enhance software development efficiency by providing real-time coding suggestions, improving review processes, and accelerating software development activities through advanced natural language processing capabilities. Similarly, there are many other ways to mash up AI and software engineering for applications, ranging from automated coding to bug detection and streamlined testing processes. Generative AI, among the various disciplines of AI, has the potential to reshape the software development landscape in ways that were unimaginable a few years back.
Key Takeaways and Webinar Recording
Check out the session recording and the key takeaways (along with timestamps in the recorded video).
The Generative AI Interaction Planes
With so much happening in the Generative AI space, it is always helpful to build a pattern that acts as the guiding principle for applying this technology to the relevant use case. One way to look at this pattern is through the lens of interaction planes.
Any Generative AI based application has four interaction planes. These can be depicted as follows.
The input and output planes are the most commonly used planes. The input plane feeds a document, text or media, and the output plane generates text, image, media or code.
But, if you have a specific request, you need the prompting plane.
Further, if you want to customize the underlying model as per a specific persona or configuration, there is yet another plane.
Together, all these four planes constitute the main levers through which you can exercise Generative AI in software engineering and control the outcome of the underlying machine learning models. This is the Gen AI user interaction architecture.
These planes constitute the key technical evaluation considerations for choosing the best Gen AI systems for software development for a specific software development lifecycle phase.
You can think of these four interaction planes as a way to realize Generative AI use cases, via interaction with the underlying Gen AI model, such that:
- 1The input plane allows users to feed in data. This may also include RAG input or external sources.
- 2The prompting plane lets the user ask specific questions.
- 3The persona plane offers configuration options for the underlying model to generate domain-specific responses.
- 4The output plane presents the generated outcome from the model.
Let’s explore the use cases and identify the Gen AI platforms and tools most suitable for SDLC.
Gen AI Use Cases in the Planning and Requirements Phase
Gen AI adds value during a software development project’s planning and requirement phase by offering insights into previous sprints or release cycles and existing requirements, helping product managers better plan upcoming releases. Additionally, the natural language processing capabilities of Gen AI aid in understanding and summarizing requirements, making it easier to capture and interpret user needs accurately.
Use Case #1 - Requirement search and summarization
This is the most obvious use case of document summarization. It is derived from the general usage of any LLM, which consumes information from a document and summarizes it based on a prompt. The output plane displays the summary and offers a search feature to summarize specific keywords in the document.
In the case of SDLC, the document is an SRS/SRD (Software Requirement Specification/Software Requirements Document) or a PRD (Product Requirements Document).
Gen AI tools for requirement search and summarization:
Advanced LLM-powered topic research engine that generates responses based on requirement docs or associated content and data for quick search and summarization.
Use Case #2 - Requirements enrichment and insights
This use case goes a few notches ahead to recommend improvements in the existing requirements. This is beneficial to:
In this case, the input plane takes in additional information from the source code repository and issue tracking systems.
The output plane is enhanced to generate improved requirement statements besides offering search and summarization capabilities. Additionally, it can generate recommendations for engineering productivity, common issue patterns, and general health of the development processes.
Gen AI tools for requirements enrichment and insights:
Requirement analysis for improving use stories, identifying missing requirements, and deep analysis of ambiguities, beyond functional scope.
Insights on engineering team bottlenecks, and monitoring of developer productivity issues based on data from Git repositories.
Use Case #3 - Requirements generation
This use case helps in generating new requirements, either from scratch or based on existing requirements. The Gen AI interaction plane is augmented with an AI persona that acts as a product manager to translate high-level requirements written in natural language to detailed product requirements for building software applications.
Gen AI tools for requirements generation:
Requirement generation, feedback, and collaboration tools to help product managers brainstorm ideas, write PRD specs, and support design efforts.
Quick PRD generation for product features based on prompts containing an overview and high-level requirements list.
Gen AI Use Cases in Design and Coding Phase
These use cases add value to the core developmental activities of software development teams.
Use Case #4 - Design artifact generation
Gen AI has the ability to create visual content. This capability can be leveraged to generate:
The interaction planes for generating such a wide array of diagrams and illustrations require several enhancements. First, a chain of thoughts or commands must drive the prompting plane. The persona plane is also a specialized model trained to generate illustrations in a specific visual format. The input plane can ingest existing code and technical manuals to better understand how to generate design artifacts.
Gen AI tools for design artifact generation:
Professional grade, UML-compliant diagram generation based on well-known models (GTP, Claude, Gemini, Qwen, and more.
A product development workspace platform for collaborating on ideas and creating diagrams and wire frames to build concise product documentation.
A tool built specifically for engineering and technical design with knowledge of cloud-specific architectural icons and deployment patterns.
A business-centric, visual collaboration tool for designing illustrations for business planning and research. (SWOT, Gap Strategy Frameworks)
Use Case #5 - Development assistance
This is one of developers’ most sought-after use cases for analyzing, searching, and explaining existing source code and performing code completion, edits, modifications, and refactoring by generating new code.
In terms of the Gen AI interaction planes, there is a significant focus on the persona and output planes. Since this use case deals only with source code, the persona must be specialized and trained in programming languages, with knowledge of frameworks and deployment setup. For example, for generating code for a Node.js, express.js application to be run on the AWS cloud, the persona must understand JavaScript, the express.js framework, and AWS-specific deployment options, such as CloudFront. The output plane is capable of generating code and comments. The comments become part of code review comments or explainer texts.
Gen AI tools for development assistance:
Code intelligence platform capable of generating insights, providing contextual search, and offering editing assistance at code level.
Coding assistant focused on generating performant and quality code, ideal for fixing and refactoring.
Complete natural language interface for translating ideas to apps with code generation capabilities for various web development stacks.
Complete platform for development, application, and infrastructure security with real-time security scans and fixes at the code level.
AI-driven code review with built-in linters and SAST tools for almost all programming languages.
A curated collection of code analysis, security coverage, and productivity tools with minimal configuration and maximum language support.
Automated code migrations and dependency management with a DSL (Domain Specific Language) for searching, querying and modifying source code.
AI-based code mentoring, guidance, and improvement for programmers of different proficiency levels.
Use Case #6 - Peer programming
Also known as “auto-complete in steroids”, this use case offers live coding assistance within the developer’s native environment.
The Gen AI interaction plane for peer programming includes a more nuanced prompting interface since the underlying model has to understand the local context of the code being written or edited by the developer. The persona plane is also deployed as a more specialized model that understands the programming language used for peer programming.
This collaboration between AI tools and human software engineers ensures that AI can provide inline recommendations to augment the problem-solving, and contextual understanding of human engineers.
Gen AI tools for peer programming:
Complete management of the development workspace with real-time recommendations on coding, PRs, issues, and rich extensions to customize the coding workflow.
Feature-rich AI code editing with emphasis on security, collaboration, and advanced troubleshooting.
Standalone IDE with realtime coding assistance with code edits, rewrites, predictions, and smart reference searches from docs and the web.
High-performance AI code editor with interactive programming features via inline REPL capabilities to visualize the logic and data.
Agentic code editor with reasoning engine to collaboratively build business logic flow and deep contextual awareness for quick refactoring.
Gen AI Use Cases in the Testing Phase
The testing phase has a universal need for test case generation and automation. With advancements in AI technology, the role of software test engineers is evolving, necessitating the enhancement of their skills in applying AI and machine learning to software test automation. Historically, this was achieved using code-level parsers to generate test cases based on the semantic understanding of the code to identify and list success, failure, and boundary cases for each execution unit of code, such as procedures and functions.
Use Case #7 - Testing
With Gen AI, test generation and automation can be further expanded to contextual test cases that depend on specific UX workflows, screen resolution (responsiveness), device configurations, and load conditions. One of the groundbreaking innovations in this space is agentic user testing, which achieves the depth of manual testing at the speed and scale of automated tests by simulating many users.
The Gen AI interaction plane for testing is very similar to the previous phases, but the output plane differs in its ability to generate executable test cases and supporting documentation.
Gen AI tools for testing:
Agentic User Testing with context-aware automated tests and exploratory UI coverage for desktop and web applications.
Extensive Gen AI test suite with support for many browsers, OSs, and devices for highly scaled-up text execution and observability.
JVM-based unit testing within IDE with continuous alignment between code and test in real-time for traceability and maximum test coverage.
Visual Studio Code extension for unit test generation, documentation, and recommendation for JS/TS/Py.
Gen AI Use Cases in Deploy and Maintenance Phases
The deployment phase involves observability and monitoring of the software application. Any bugs discovered in this phase are channeled through the DevOps process as part of the perpetual maintenance cycle. Gen AI finds applicability in both phases.
Use Case #8 - Autonomous Observability
Autonomous observability allows developers to perform root cause analysis of issues through intelligent log analysis and debugging. In some ways, this applies to the coding phase, as debugging is also performed as part of unit testing.
The Gen AI interaction planes for this use case can ingest a huge amount of log data as part of the input plane and generate recommendations for pinpointing the exact source of issues and critical bugs as part of the output plane.
Gen AI tools for autonomous observability:
Log intelligence from applications and infrastructure for improved ITOps with automated issue workflow management with AI.
Built-in observability IQ with root cause analysis for complex debugging scenarios with less noise and faster issue detection and resolution.
Autonomous debugging within the IDE, from analyzing issues to pinpointing the code along with logs and snapshot recommendations for production environments.
Development team collaboration with built-in AI assistant for streamlining debugging with automated source code analysis.
Use Case #9 - DevOps enrichment
This use case extends the autonomous log analysis to recommend DevOps process improvements based on continuous assessment of the health and performance of runtime applications and delivery pipelines.
The Gen AI interaction planes have the same capability required for autonomous observability. However, the input plane is augmented with DevOps logs along with the software application’s runtime logs. This can include build logs, application environment, and access configurations.
Gen AI tools for DevOps enrichment:
DevOps CoPilot to combine predictive and causal AI and learn from configuration, observability, and security data for AI-powered insights.
A suite of tools and integrations for 360-degree DevOps monitoring and automation in a platform-agnostic way.
Gen AI Platforms and Tools Applicable for SDLC
General Guidelines for Gen AI Adoption in the Software Development Lifecycle
Narrowing Focus: Gen AI has the potential to supercharge many aspects of the SDLC phases. However, the focus should be on applying Gen AI to gain productivity and performance boosts in any one aspect at a given time.
Privacy First: The confidentiality of source code and documentation takes priority over everything else. Hence, the Gen AI platforms integrated with a project's source code and docs repository must be certified for SOC2 Type II.
Maturity Alignment: The Gen AI intervention must align with the maturity level of the team or organization to avoid a sprawl of AI-generated content without human oversight. In the worst case, this may lead to a chunk of AI-generated code that finds no takers to fix bugs later.
Non-Functional Matters: Using Gen AI, it is best to prioritize non-functional aspects of the source code, such as security, scalability, and redundancy, over functional aspects. Developers tend to focus more on functional aspects, while non-functional aspects are overlooked due to the cognitive overload of developmental activities and shorter sprint schedules.
Domain Check: Every Gen AI solution adopted for software development must be checked for domain-level specialization and understanding of the AI persona. This includes the programming language, deployment architecture, frontend, backend, and middleware-specific platform considerations.
No Gospeling: All Gen AI platforms and tools must undergo regular internal audits to ensure the generated output aligns with the product and business objectives and to avoid blind assumptions or acceptance of the generated outcome.
