Generative AI can speed up coding tasks, but developers using AI need to keep some risks in mind, according to a June 2023 study from McKinsey. Recent studies have found both positive and negative impacts of generative AI for software developers. Technology leaders can help their teams reduce security risks from generative AI while getting the most out of generative AI at the same time.
This article is sponsored by UST, which offers enterprise cloud services to help transform organizations’ cloud strategy, governance models, cybersecurity and application development.
Pros of generative AI for developers
McKinsey found that using generative AI sped up coding in a variety of situations. The report doesn’t specify which company’s AI was used, but says the study used both prompt-based natural language generative AI tools and generative AI tools trained specifically on code base and embedded into a developer’s integrated design environment. Task completion on relatively simple tasks coders might encounter in their everyday work (documentation, code generation and refactoring) could be performed 20% to 50% faster with generative AI.
Generative AI was particularly good at the following:
- Repetitive and manual work
- Creating a first draft
- Updating existing code
- Explaining new concepts, providing step-by-step guidance or helping developers gain a basic knowledge of a code base, language or framework
- Letting developers focus on meaningful work and get into a flow state instead of spending their time on menial tasks
Code quality in relation to bugs, maintainability and readability improved with generative AI’s help. However, participants noted that they iterated along with the generative AI instead of taking the exact content it produced. AI is best used to augment, not replace, developers, McKinsey stated.
SEE: This cheat sheet explains what generative AI does and how businesses can get started using it. (TechRepublic)
“Ultimately, to maintain code quality, developers need to understand the attributes that make up quality code and prompt the tool for the right outputs,” McKinsey authors Begum Karaci Deniz, Chandra Gnanasambandam, Martin Harrysson, Alharith Hussin and Shivam Srivastava wrote.
A 2023 academic study cited by Brookings found professional programmers writing an HTTP server in JavaScript using the generative AI tool GitHub Copilot took 55.8% less time than a control group performing the same task without generative AI. However, there was no significant impact on the success of the task.
Cons of generative AI for developers
More complicated coding tasks didn’t see the speed boost from generative AI that simpler coding tasks received. Time savings lowered to 10% on tasks developers deemed difficult, such as working in a programming framework they weren’t familiar with, McKinsey found.
Generative AI didn’t help less experienced developers complete their tasks more quickly. In fact, some tasks took junior developers 7% to 10% longer with generative AI assistance.
Three areas that required human oversight were:
- Debugging and finding errors
- Working within organizational context, such as who the code will be used by, the other interfaces and systems the software will interact with and what data the software will use
- More complicated coding requirements, such as combining multiple frameworks with disparate code logic
“The metrics today around generative AI talk about the percentage of generated code being utilized,” UST Senior Director, Cloud Practice Dan Bode told TechRepublic. “Today it’s just an accelerator, but still requires an experienced developer to ensure that it creates logic that works as a part of a larger system.”
How to reduce security risks around generative AI
To reduce risks, organizations using generative AI for work should be sure the AI model doesn’t feed proprietary data back into training sets. Developers should be aware of potential regulatory or intellectual property problems that may arise from using generative AI, McKinsey said. They should double check recommendations given or code written by generative AI.
Any AI model used for work should be designed in accordance with the organization’s overall goals, including accountability and security, which can be critical in software development.
However, security is a place where generative AI could save time. Security is one of the top two tasks developers spend the most time on, according to a June 2023 study from GitHub on generative AI use among 500 U.S.-based developers at enterprise companies.
How technology leaders can help their teams in the age of generative AI
In order to get the most out of generative AI in coding, IT leaders can deploy exercises and training workshops. In particular, it would be useful for the team working with generative AI to understand prompt engineering and potential copyright issues. Developers using generative AI should be aware that generative AI doesn’t always give correct answers and should be able to spot errors. Tech leaders may need to be prepared to shift staff to higher-value tasks if generative AI use leads to additional capacity, McKinsey said. Technology leaders should be aware of risks including privacy, security, regulatory changes and ethical issues.
“AI can help us generate ideas, but it is up to us to decide how to use them,” wrote Priyanka Vergadia, staff developer advocate at Google Cloud, in a blog post on May 26.
Vergadia pointed out that low code/no code AI platforms can help highly-skilled technical employees automate some of their simpler tasks. Technology leaders might want to think about areas where highly-skilled and nontechnical employees may be able to work together more smoothly using low code/no code generative AI.
About the sponsor
UST cloud service teams help their customers realize the value of the cloud. Services offered include:
- Design: UST can help develop a roadmap for innovation based on data-driven insights that balance speed, cost and security.
- Implementation: UST excels in operationalizing and automating cloud initiatives, leveraging our ecosystem of services, technology building blocks and industry partnerships to accelerate time to value.
- Governance: UST offers managed services, audits and a deep commitment to ongoing collaboration help organizations to adapt and grow alongside UST.
UST’s customers are among the largest brands across industries such as healthcare, finance, retail and more.