Get Started
AI will unlock developer productivity

ARTICLE | March 11, 2024 | VOICES

AI will unlock developer productivity

The future of programming will pair human coders with AI assistants, reducing repetitive, boring tasks and maximizing creativity and problem-solving 

By Anna Byers, Carla España Lynch, and Tanuja Sawant, Workflow contributors

Generative AI’s (GenAI) explosion onto the technology scene has created a world of possibilities for improving productivity across countless industries and roles. Application development is a particularly exciting space for AI. This is true for two reasons. First, there is a shortage of human developers, which means that demand from businesses on their IT development teams far outpaces what can be delivered. (In response, Gartner predicts 80% of low-code app development will happen outside IT departments by 2026.) Second, research shows that most developers feel motivated in their jobs when they can spend their time solving new problems and operating quickly. GenAI has the potential to address both of these issues by accelerating development and boosting productivity for no-code to pro-code developers.  

To unlock these productivity and scalability benefits, it is essential to understand how developers perceive both the benefits and risks of AI in software development. Through our own extensive research, we have gained the following valuable insights into how best to support AI-assisted software development.

Report

Future proof your workforce

Read now

We are seeing a strong appetite for AI-assisted development because of promised increases in productivity and efficiency. Of the 254 developers across companies we surveyed, 95% are already using some form of AI in their work, and 86% indicated that they are excited about AI’s potential to make them more innovative and productive.  

These findings are not unique to our survey. According to a 2023 GitHub survey, 92% of U.S.-based developers reported using AI for both work and personal projects, and 70% saw a significant benefit from using it. A study by McKinsey similarly found that 87% of developers felt that GenAI allowed them to focus on more meaningful work.

Our research has consistently demonstrated that a developer’s skill level impacts how they perceive AI. Highly skilled developers are more likely to already be using and benefiting from the technology, increasing their trust and confidence in the value of AI tools. Both highly- and mid-skilled developers see concrete benefits, such as reducing time spent on repetitive tasks. At the same time, no-code and low-code developers are more wary, expressing concern that they don’t understand how AI works and don’t know how to identify and fix errors it might produce. Helping them understand the benefits and limitations of AI, and how to use it, will boost their confidence (more on that later).  

Across all these groups, but especially low-coders, our research has shown that AI is seen as a valuable learning tool, helping developers gain new skills and remember forgotten syntax. And, according to Stack Overflow’s 2023 Developer Survey, 55% of those who are just learning to code are using AI in their development process. 

While accuracy is important, AI recommendations don’t need to be perfect. Developers say they are comfortable with tools that generate inaccuracies if they can easily be addressed, provide a useful launchpad for developers so they aren’t starting to code from scratch, and increase overall human efficiency. 

From our research, it’s clear that developers feel AI output is merely a starting point to build from and that they need AI-generated recommendations to be contextually aware and consistent. Developers have strong requirements to be able to review, modify, and test AI-generated outputs to protect the security of their platforms. AI can be of use here by generating recommended tests and scanning code for errors, thereby reducing time spent manually testing code, a challenge that developers repeatedly face. 

To get the most value from AI, developers would benefit from clear guidance, best practices, and user control. Our research has shown that developers come to AI with preconceived notions about how it will learn from them, make specific recommendations, and deliver accurate results. In order to ensure that developers have a positive experience and keep using AI in software development, transparency and “explainability”—AI’s ability to clearly explain how it operates—are a must. 

Another critical area is best practices for prompting GenAI models. We hear from developers that they need concrete examples and recommendations for how to get the best output. Prompting is a new skill set for many developers, and to be successful, they need to learn how best to create and refine the most effective prompts. 

At the same time, user control is essential. Developers are comfortable with some degree of error as long as they can accept or ignore recommendations (commonly referred to as a “human-in-the-loop”). Given the iterative nature of coding, developers need AI systems to promote experimentation and flexibility, enabling humans to iteratively adjust prompts and explore additional capabilities of the technology. 

 

The potential of AI to transform development is clear, but what must not be forgotten is that the best technology in the world is useless if people refuse to use it.

Lastly, but perhaps most importantly, it is crucial that developers have a high level of trust in these systems. According to our research, developers cite privacy, security, and accuracy as their primary concerns with AI, making transparency key to drive future adoption. 

Security risks are of especially great concern, since the code output from AI can incorporate vulnerabilities present in the code that was used to train the AI system. Therefore, it's imperative that businesses apply consistent security and governance policies to AI-generated code, treating it with the same scrutiny they do human-written code. 

ServiceNow has developed its own set of human-centered AI guidelines (soon to be publicly available) that allow us to build trustworthy and responsible AI. Additionally, we recently joined the AI Alliance, an international community of leading technology developers, researchers, and adopters pushing for open, safe, and responsible AI.  

Our research has shown how to best customize AI tools to provide the best experience for human developers at all experience levels. With the support of—and trust in—AI, human developers will soon realize huge productivity gains, turbocharging development. Seasoned human developers collaborating with AI will work faster than ever and be able to spend more time on the thorniest and most interesting problems. With AI-assisted testing, code will be cleaner and more stable. For junior developers, AI will serve as a trusted learning partner, offering recommendations and accelerating professional growth. For those without coding experience, low-code development supported by AI will reduce barriers to entry, taking a user’s text input and turning it into an application in minutes.  

The potential of AI to transform development is clear, but what must not be forgotten in the rush toward this AI-enabled future is that the best technology in the world is useless if people refuse to use it. For this reason, we must design these tools to work best with—and cater to—humans themselves. If we do this right, it will improve the work (and lives) of not only developers, but anyone who works with a computer —pretty much everyone—and help us to realize  the true promise of digital transformation.

95%
of developers surveyed are already using some form of AI in their work.

Trending articles

Related

GenAI Makes Coding and Automation Accessible to More Users. Learn How

Get ebook

Related articles

The Low-Code Innovators
ARTICLE
The Low-Code Innovators

Engineers aren’t the only ones creating sophisticated apps. Meet four citizen developers whose handiwork has made a difference. 

Year two of the AI revolution
ARTICLE
Year two of the AI revolution

If 2023 was the year of experimentation, things get real in 2024. Here’s what the experts expect next year 

Want to build apps fast? Put AI to work
ARTICLE
Want to build apps fast? Put AI to work

How open-source, open-governance AI models could revolutionize software production 

Author

Anna leads UX Research for ServiceNow’s Developer Experiences
Anna leads UX Research for ServiceNow’s Developer Experiences. She has conducted research across enterprise, consumer, and academic settings for 15 years.
Articles by Anna Byers

Author

Carla España Lynch, PhD, is a Staff UX Researcher on ServiceNow’s AI Platform team, where she advocates for human-centered AI.
Carla España Lynch, PhD, is a Staff UX Researcher on ServiceNow’s AI Platform team, where she advocates for human-centered AI. She has conducted research for 15 years, spanning industry and academic domains. 
Articles by Carla España Lynch

Author