Generative AI meets software development: The advent of generative coding

HFS predicts the rapid rise of “generative coding,” which goes beyond the current use of generative artificial intelligence (GenAI) to convert code. Generative coding will embrace the automated creation, manipulation, conversion, and optimization of code. While humans will oversee the requirements and approve the delivery of outcomes, generative coding will intertwine human and machine coding practices throughout the software development lifecycle (SDLC) to craft composable applications and microservices that human and machine teams reintegrate as microservices—built for cloud deployment and curated by Kubernetes.

Artificial intelligence (AI) is taking off because employees need to understand and action large amounts of data at record speed in natural language (all languages, not just English). The breakthroughs we see now result from the cloud bringing near-infinite availability, storage, and computing—resilient architecture at a global scale. As a result, the internet, data, and AI technologies collided.

Exhibit 1: The internet, data, and AI are like three galaxies colliding—their shock waves will destroy and create at an alarming rate

Source: HFS Research, 2023

This collision is triggering dramatic changes in software development. The principal changes will be in code assessment, development, conversion, and testing. By adding generative GenAI tools, the future of programming will require the human software engineer to operate with new team mandates.

Each software development team member will have multiple machine and human collaborators, and their teams’ success will be based on how they manage the capabilities of these resources. Software development leaders will need to coordinate teams of people with GenAI tools, which use probability algorithms to understand the business and technical intent of the code and refactor it into desired computer languages. A human alongside or in the middle will be essential to improving the code and how the machine learns and improves its own coding skills.

Generative coding is made possible by the emergence of the new S-curve of value creation

The S-curve of the Generative Enterprise™ leads to a fundamental shift in many parts of our business. Exhibit 2 illustrates that we are at a tipping point where IT and business services are transitioning from a global enterprise to a generative (AI-driven) one. We think a major point will be how software is developed, optimized, and engineered. Akin to the S-curve of value creation, in the new S-curve of software development, we’ll find ourselves involving teams made of man and machine.

Exhibit 2: The Generative Enterprise inspires a new S-curve of value creation

Source: HFS Research, 2023

Using machines to augment creating and refactoring code will free teams from low-value tasks and allow them to take on more abstract work based on inputs from projects, business, and customer requests. The result is more valuable work for employees, and a massive amount of software will be created with this new model.

The software innovators’ dilemma isn’t about speed, creation, or capacity—it’s about quality

We are at a new juncture in software development. We will no longer have specialists in specific code platforms; machine-based AI coders will make that irrelevant. Instead, we’ll have coders that operate as artists, using code and AI co-pilots to develop solutions from existing and brand-new software development practices.

Currently, generative coding experiments focus on converting legacy code by using AI to recognize code logic and, in some cases, converting one type of code to another, for example, converting COBOL to Java or node.js. As the code generator understands more of the logic needed for both the original and target codes, its ability to automate conversion improves. While current tests are for nowhere near a 100% conversion, they free resources to focus on more complex coding and developing and testing code in new projects the business and technology teams require.

This generative coding concept is only the beginning. In a coming study, HFS will postulate how generative coding will evolve the SDLC (software development life cycle) and the role of software development. We’ll examine how these tools continue to emerge, what they can do today, what is possible, and what might be next.

HFS guidance on adopting generative coding as part of an enterprise SDLC

Embracing this change in software development is expected to take many forms and likely result in a few missteps, but sitting on the fence isn’t an option. We recommend the following as a starting point:

• Consider participating in an industry forum about certifying generative AI models for coding. With a push for ethical coding and to ensure code is crafted to be supported, maintained, and improved as newer technologies or business needs arise, we’ll need a measure of trust in place. Implementing a certification program for bots is the best way to accomplish this. These certifications should include human-in-the-loop clauses to ensure bots submit to peer review and quality assurance (QA), requiring the same rigor we do from our organic coders.
• Encourage dialog about how your software development teams will need to embrace machine co-pilots to augment and accelerate coding. We are at the juxtaposition of multiple coding platforms radically converging around business logic, incorporating data in new ways, and using the cloud as a home base for all development. This will be the new reality; begin developing skills for this today.
• Recognize this moonshot as an ecosystem play. Engage your software vendors, services partners, and business stakeholders to craft the questions about how adopting generative coding can lead to measurable, sustainable business outcomes.

The Bottom Line: Generative coding is emerging as a trend to facilitate velocity in the SDLC. While it still requires significant oversight and quality control, as these models become more adept, enterprise leaders will need to pivot how they resource all aspects of the software development value stream.

One of the first areas expected to experience a significant impact from GenAI is software development. Companies, software development firms, and ISVs have already begun experimenting with how GenAI can speed code creation and conversion. As the internet, data, and AI technologies collide, enterprise CIOs and applications and software teams need to start preparing for how they will manage their talent and SDLC process. HFS highly recommends preparing your software development, quality assurance, testing, and applications support teams for the advent of generative coding.