Implementing comprehensive AI solutions across the entire development lifecycle is crucial for maintaining competitive advantage and maximizing developer productivity.
Since the emergence of AI-based coding assistants, enterprises have rapidly adopted these tools, making them a strategic priority for CIOs. While developers are familiar with assistive technologies like IntelliSense and advanced debugging in modern integrated development environments (IDEs), AI coding assistants promised significant productivity gains that attracted immediate attention.
Over the past year, organizations invested heavily in training programs for these coding tools and began implementing them in real projects. However, many discovered that the expected productivity gains weren’t fully materializing, prompting a deeper examination of their effectiveness.
Study reveals coding assistants’ limitations
TCS recently conducted a systematic study of over 40 programming assignments across varying complexity levels within an insurance company’s application portfolio.
Developers were given access to coding assistants and initial training. With expert guidance on effectively using these tools, we tracked improvements in both productivity and developer satisfaction. Several key insights emerged:
- Most work involves enhancing existing code (brownfield development) rather than creating new applications.
- Developers spent over 50% of their time on “impact analysis” — understanding poorly documented code and determining what needed changing.
- Coding assistants excelled at generating individual code snippets but struggled when integrating different components (UI, API, databases).
- These tools performed well with modern languages like Java and Python but were less effective with legacy systems using PL/SQL or Unix scripts.
- Developer satisfaction improved dramatically, with 92% reporting they produced better code or achieved higher productivity.
The message is clear: while coding assistants improve the coding process, they address only part of the software development lifecycle (SDLC). Enterprise needs vary widely in coding standards, architectures, and developer skill levels, requiring more comprehensive solutions.
Beyond coding: Requirements for comprehensive AI development solutions
A complete AI platform for application development should include:
- Software change impact analysis
- AI-driven re-engineering from legacy systems
- Business functionality documentation
- Modern user experience (UX) design while intelligently reusing existing layouts
- Starter prompt kits for new engineers
- Integration with enterprise coding standards
Solutions like TCS’ AI Catapult Forward address these needs by enabling end-to-end AI capabilities throughout the software development lifecycle (SDLC). Built on Microsoft Azure Cloud and compatible with GitHub Copilot, this platform provides features such as AI-powered impact analysis, software re-engineering from legacy systems, and intelligent documentation of business functionality — all while respecting each enterprise’s unique coding standards and architectures.
Enterprises need more than just coding assistants because successful application development requires addressing the entire software lifecycle, from understanding legacy code to integrating components across different systems. As enterprises continue evolving their application portfolios, implementing comprehensive AI solutions across the entire development lifecycle will become increasingly essential for maintaining competitive advantage and maximizing developer productivity.
To find out more about TCS’ AI Catapult Forward click here.
