AI Code Completion: A Rising Trend

As I will happily demonstrate, it doesn’t take a genius to point out that artificial intelligence (AI) is a very, very hot topic right now. In fact, G2’s 2023 Best Software Awards showed that the top 3 fastest-growing software products on G2 were AI tools. Much of the buzz is around Chatbots, G2’s fastest-growing AI category, with a staggering 261% traffic increase between February 2022 and February 2023.

However, AI's current and future impact on software development is a major part of the conversation. AI code completion, a type of generative AI which can range from one-line code suggestions to fully functioning programs based on codeless prompts, represents a leap forward for the entire development process. 

The buzz around code completion 

Recently you may have heard people saying things like “AI will replace developers” or “AI will usurp humanity and rule the world via an uncaring mega network.” You’re on your own regarding the latter, but the former won’t be happening anytime soon. 

Unlike some other forms of generative AI, such as synthetic media, code completion can’t go from prompt to finished product without significant human oversight and intervention. For example, you can prompt an AI image generator to make a picture of a funny horse and wind up with a unique image that looks exactly like a funny horse should. AI-generated code, however, must be checked against security, quality, and compliance standards before it can be used in a production environment. 

Because programming is a complex and inherently creative process, it’s unlikely that companies will be comfortable with unsupervised AI generating their code. 

Rather than focus on whether AI could ever truly replace developers, it’s perhaps more useful to consider generative AI as a partner during the development process. Code completion tools like GitHub Copilot and Amazon CodeWhisperer position themselves as exactly that. These tools use machine learning to produce code suggestions and fill gaps in existing code based on what’s already there.

The result is something far less frightening than AI replacement: speed. Compass OUL recently released a study that showed a predicted 54% decrease in the time it takes experienced developers to complete tasks when comparing those tasks without AI to the likely state of AI in 2025.

This will ideally allow developers to (checks notes) “automate tedious tasks and spend more time on complex projects.” Of course, that phrasing has been around as long as low-code, and it usually doesn’t mean anything substantive. I think that the prospect of finishing tasks in half the time it used to take is a good enough talking point.  

Code completion is being mentioned more on G2

We already know that AI is buzzing on G2. But does that extend to the development space? At the time of writing, G2 hasn’t launched its AI Code Completion category, but reviewing the data shows a clear trend.

From Q4, 2022 to Q1, 2023, mentions of code completion across G2 reviews site-wide have increased 182%. This follows a general decline in mentions during 2022, so AI code completion has clearly come back into the spotlight as part of the recent AI buzz.

Code completion is not without its challenges

Buyers should use caution as they adopt code completion tools. Speed and efficiency are always great, especially under the supervision of an experienced developer. But this is a rapidly growing space, with tools like Salesforce’s recently unveiled Einstein GPT able to generate more and more code with only a prompt or a few parameters. 

As non-developers gain the ability to fabricate more lines of code without knowing how that code actually works, serious questions emerge. 

How will this impact bug tracking? With no actual author to look to when things inevitably break, the debugging process may slow down in a way that significantly offsets the benefits of code completion. Security compliance is also a significant concern; code completion tools must be able to guarantee best-in-class security standards to avoid cybersecurity nightmares. 

Looking forward

The AI trend is exciting, scary, and perhaps even a bit annoying, depending on who you’re asking. In the software development world, people have already begun to question whether developers’ jobs are on the line in the near future due to generative AI.

The more likely outcome, however, is that code completion will continue to serve as a major benefit to developer productivity. Rather than have their jobs replaced, developers will find themselves better equipped to do their jobs.

While all the advantages come with concerns, only time will tell whether innovation in this space will outpace companies’ abilities to avoid major debugging and security headaches. Look out for G2’s soon-to-be-released AI Code Completion category to keep an eye on this space.

Edited by Shanti S Nair