The emergence of AI has sent shockwaves through many industries, with software development having a front seat.
Artificial Intelligence (AI) has been making headlines in recent years and is set to become a major disruptor in the job market. In the publishing industry AI has already shown it can be disastrous for copyright and poses an existential threat for writers and those working in the publishing industry. No industry is immune to the effects of AI, with software development seeing exponential adoption of AI tools. Running from GitHub CoPilot to Codeium, who make writing software easier than ever before.
GitHub CoPilot
In June 2022, GitHub launched its AI coding assistant GitHub Copilot, which the platform heralded as the first ever AI pair programmer. The tool used by AI can fully write and complete code. In June 2023, Freethink spoke with GitHub’s CEO, Thomas Dohmke, about the impact of AI and its Copilot solution on software development. Dohmke predicted that systems that would break down complex software problems into smaller problems, noting that since the punch card, code has become ever more complex, featuring trillions of lines of code.
This has resulted in developers having to work with complex and vast coding sets, which require a massive amount of skills. AI programs, Dohmke notes, can break down these large data sets into digestible crunch. However, it will take multiple iterations before developers can go hands off and leave source code reviewing to AI tools.Dohmke observes that developers aren’t necessarily coding as much as they used to, noting that two to four hours a day is spent on writing code.
The other time is used for stand-up meetings and analyzing crash reports, referring to App Store Reviews. CoPilot can help improve and reduce the time spent coding. As districations pile up, CoPilot helps developers remain focussed. The CoPilot program started in August 2020 Dohmke explained, seeing its first internal preview in the early stages of 2021, where 250 GitHub employees could play around with the new tool. The response was overwhelming. Employees were excited and made CoPilot a part of their daily routines.
The NPS score improved, coming from a minus 100 to plus 100, with testers giving an NPS score of +73. No other change in the development environment gave such promising results, Dohmke noted. Developers kept using the tool, with retention numbers staying stable as time progressed. CoPilot saw exponential adoption within GitHub and its parent company Microsoft. When asked how CoPilot contributed to real life efficiency gains, Dohmke noted that the amount of tab key strokes reached an average of 46 percent.
As time progressed, 80 percent of code was written by CoPilot. CoPilot allows developers to make their style more explicit, with the tool learning how it can add and improve the code, synthesizing new lines. A quantitative survey among 2,000 developers revealed that of the developers who used CoPilot, 75 percent felt more fulfilled in their jobs. Writing code required less mental capacity and less time was being spent on boilerplates.
In April 2023, CNBC reported that disruptions were underway for software developers and engineers. While ChatGPT was making waves among the general populace, tech workers were already anticipating massive disruption, with some seeking new opportunities or adding AI to their skillset. Chief executive officer at Ceipa, Sameer Penakalapati, explained it’s unfair to attribute complete job abolishment to the emergence of solutions like ChatGPT, however, tasks and responsibility might be reshuffled.
Penakalapati added that GPT programs can generate codes at much faster rates than a coder or tester, minimizing the code creation time from days to week, down to several seconds. Much faster than any human could ever do. Herein lies the prowess of mathematically powered GPT programs, who are capable at pattern recognition. Generative AI meanwhile can impact the jobs of software devops in multiple ways, ranging from development and code deployment to maintenance. These tools won’t be able to replace humans completely, Penakalapati notes.
The number of developers and engineers required to create software will decrease, he warns. Despite the reduction in the developers required, across the entire development process, from writing, to testing to designing, human intervention will be necessary. Partner at the McKinsey Global Institute, Michael Chui, adds that AI will impact every knowledge worker. Noting that technologies such as GPT, can create first drafts rapidly, ranging from writing, generation code, to more creative elements such as images, video and music. Chui adds that at some point in time, workers will forget how it was to work without AI, comparing an era when Excel and Word became irreplaceable tools for the workplace.
Google’s view on AI
In May 2023, Staff Developer Advocate for Google Cloud, Priyanka Vergadia, detailed the company’s vision of the impact of AI on software development. Google is caught in the middle of the AI revolution, releasing its own products that have to compete with upcoming tech giants such as OpenAI. Vergadia notes that AI won’t displace all technical jobs. While these tools are capable at automating a lot of menial tasks, AI won’t be replacing creativity, intuition and problem-solving abilities of human developers.
However, AI will allow for greater access to those less experienced with data science. This is a point that was also brought forward by CEO of Codeium, Varun Mohan, later in the article, where AI can enable a wave democratization. Allowing for more people to enter the field as less expertise is necessary to hone the craft of software development. While an understanding of data science is necessary, developers can tap into pre-trained models to build their own software experiences. The usage of pre-trained AI models will decrease the cost of training AI models, as they can serve as a foundation model, Vergadia argues.
Cloud-based models can bridge the starting and growth phase, giving developers scalable infrastructure and pre-built frameworks through which they can expand their solutions. The impact of AI is already undeniable, with its solutions having disruptive effects across many industries, ranging from healthcare, finance and many more. Vergadia notes that AI will require human oversight, despite its capabilities. While this might be true a wave of uncertainty has crept among many developers.
Developers fear AI
A December 2023 Motherboard poll, among 9,388 engineers, in cooperation with Blind, felt that the rise of AI will result in job displacement. Only a minor 6 percent was confident they could find another job with equal pay. Software developers saw major lay-offs and a tightening labor market, with problems now being compounded with the rise of artificial intelligence. The poll revealed that nine in ten surveyed software engineers felt that finding a job had become more difficult, with 66 percent stating the search had become much harder.
An overwhelming 80 percent indicated that the job market had become more competitive over the last year. The sentiment isn’t surprising when placed in the broader tech sector context, which saw over 400,000 lay-offs over 2022 and 2023, Vice noted. The changing dynamic is in stark contrast compared to the hype around coding and its subsequent degrees, which saw increased popularity in recent years. Tech-giants like Google used everything in their arsenal to attract talent, paying royal salaries, accompanied with generous secondary benefits. However, even Google was not immune to market and shareholder forces.
The techgiant has tried to find its footing among the AI disruption and has trouble keeping talent onboard despite royal pay. With the inability to find stable jobs, undergraduates are burdened with high amounts of debt. This has been a massive shift from earlier years, where recruiters were aggressively hunting for talent. An undergraduate from University of Pennsylvania, who graduated in 2010, told Motherboard there was so much competition, turning the job market into a completely different landscape. Adding that he regrets his decision.
The feeling however might be unjustified, argued Maayan Manela at Calcalistech back in July 2023. The junior position feels the most fragile, as little experience is needed to fulfill a job, making it seem easily replaceable. AI programs meanwhile have become very capable at creating stunning graphics, written content and like CoPilot, generate code. This rapid development in AI technologies sparks fear among those who serve as the first line workers and need these tasks to expand their skills set.
However, Manela notes, AI software, instead of displacing juniors at the workplace, can serve as a valuable tool to adopt new skills and reduce training time. Dr. Rafi Yoeli, lecturer in artificial intelligence at the School of Continuing Education at the Technion, told Calcalistech that people are the engine behind a company. Juniors serve a crucial role within the workforce, bringing in new knowledge and energy. But, they lack the necessary skills to develop products on their own and need years of training to work wholly independently. Companies can use AI tools to accelerate this process, by preserving and sharing knowledge.
Good tools can incorporate 70 to 80 percent of the knowledge from past employees. This doesn’t mean no job displacement will occur. In May 2023, CEO of IBM, Arvind Krishna, told Bloomberg, referenced by Reuters, that the company would put in place a hiring freeze for around 7,800 roles as it expects AI would replace these positions. The positions are primarily situated in back-office functions, such as human resources, Krishna explained. IBM estimated that 30 percent of non-customer-facing positions could be replaced through automation and AI within the next five years.
Streamlining processes with AI
In March 2024, Andrea Siedsma at the University of California San Diego questioned whether AI would replace programmers. In a collaboration with UC San Diego instructors Norman McEntire and James Gappy, they explored the potential impact of AI on the future of coding and whether the technology would wholly replace developers. AI, in its current form, was still plagued by “hallucination”, meaning the programs can generate misleading or inaccurate models, which raise questions about the programs reliability and accuracy.
Despite these shortcomings, AI tools can offer tremendous advantages for coders. Instructor for Extended Studies, Norman McEntire, commented that machine learning and artificial intelligence have been the most disruptive forces in computer software. These technologies can have a significant impact on a programmer’s workflow. Adding that the possibilities are vast and exciting, ranging from automated code generation to intelligent debugging.
As exciting as these opportunities are, having a strong understanding of coding is required to use AI effectively. Programmers will have to master the skill of conveying the role of AI on the software to their relevant stakeholders. This can be achieved by streaming routine tasks through automation. In this scenario, AI becomes a collaborator during the coding process, instead of an autonomous entity that replaces the developer.
AI supported algorithms are an example of automated programs that can result in large leaps in efficiency for programmers, Siedsma pointed out. These models scan through large datasets and offer improvements to make existing code more efficient. AI can scan code and identify bugs. Programmers in turn will transform into strategists that use AI to develop stronger, more reliable programs, Siedsma argued. A developer can complete more tasks within a shorter period of time.
Developers in turn should embrace this shift starting by using tools such as Google Gemini or ChatGPT, exploring the programs’ output through simple prompts, McEntire says. Former programmer at Sony Playstation, Hitachi and FICO, among others, James Gappy, added that students could prepare for AI by fulfilling assignments on their own and then request an output from AI programs to compare the results. This helps students get familiar with AI tools, with the running and compiling remaining the responsibility of the student.
Gappy’s outlook is more optimistic than that of those working in the field during the Motherboard survey. Gappy argued that AI will not replace programmers, however, software developers will have to future proof their careers and adopt a continuous learning mindset. Being open to collaborating with AI programs is an important pathway to improving and completing one’s skillset.
Applications from scratch
In March 2024, Senior vice president of IT and CIO at Freshworks, Prasad Ramakrishnan, explained to CIO Magazine that AI has opened new development pathways, with AI serving as a deskside colleague one could brainstorm with. At Freshworks, generative AI allowed the company to drastically reduce its development cycle. Development cycles have been reduced from weeks to days at the customer service software provider. Through a strict quality assurance and verification process, Ramakrishnan adds, the adoption of AI across workflows has been a tremendous success.
Nitin Tandon, CIO at financial service firm Vanguard said that generative AI tools have enabled tremendous productivity gains for developers. The field remains a field for experimentation he adds. New techniques are being tested to boost so-called content development and knowledge management to increase operational efficiency. Vanguard, Tandon notes, has seen positive results from its AI pilot program, which has enabled code generation, improved debugging and improved code consistency.
Within Vanguard, junior developers have seen the most promising boosts in productivity. However, Tandon remarks that the programs are still in the active research and experimentation stages. Vanguard is exploring large language models across its code base to further accelerate productivity for data engineers and developers, which in turn will generate more value to its customers.
FinTech organization Momnt has been embracing the usage of Github CoPilot. Its software development teams, from engineering to quality assurance (QA) have used generative AI solutions such as CoPilot to create test cases, with its QA teams finding new tools to conduct evaluations without losing time. President and CTO, Brian Lanehart, told CIO Magazine that the company wants its software developers and engineers to create new test cases.
Adding that Momnt has seen its teams embrace AI and new tech solutions, with employees remaining curious to test out new tools in their day-to-day process. This improves worker productivity, Lanehart observed. Software developers use generative AI to create entire applications by generating the necessary code, which significantly reduces development time. That time can be used for strategic and creative work to develop new products or contribute to existing projects.
Enabling employees
In May 2024, Techstrong spoke with CEO of Codeium, Varun Mohan about the impact of AI on application development and the job market. Mohan explained the rise of AI across development should be seen as the expansion of coding languages. These coding languages make writing code easier and more developers are now actively working in the field, with AI being another form of technology creation. Providing them the tools will result in increased adoption and new technologies.
Mohan adds a distinction will remain amongst developers, with those wanting to truly understand the core components of a system. This mindset is scarce, but also unnecessary in relation to programs such as Python. Artificial Intelligence can however serve as a bridge and decrease the barrier to entry to those who were reluctant to study and write code. AI will remove a level of abstraction, with code being more efficient. A reframing will occur, where developers will be able to focus on the core problem instead of writing countless lines of code.
Currently, developers use coding assistants, such as CoPilot, which run separately from each other. This a glorified version of hiring somebody to write the code, Mohan explained. The big shift will be artificial intelligence programs aiding developers in finding the root problem and distill the core issues into a digestible, applicable solution. AI will increase the rate at which code can be reviewed, tested and deployed, significantly reducing the time writing code.
Accelerated code generation doesn’t mean that workflows will be drastically overhauled. While code writing becomes more efficient and much faster, the rate at which developers can consume code, will not change. Mohan says that having ten times more applications, will also result in 10 times more bugs. Therefore the software development process will have to be optimized to manage these leaps in efficiency. Backend developers will need the necessary tools to be able to deploy all the newly generated applications.
If those aren’t provided, the rate of deployment will remain the same. In order to offset the need for development and deployment, Mohan argued that artificial intelligence should be able to understand the processes within a company to reach its full potential. Adding that a lot of companies are struggling to deploy all the projects they have planned. AI will help to find these bottlenecks and streamline them.
Productivity versus creativity
The verdict as to whether artificial intelligence will disrupt software development is yet to be cast. Many working in the field are reluctant and pessimistic, fearing massive job displacement. Companies like IBM have already demonstrated how AI can disrupt job opportunities. On the other hand of the spectrum are the optimists who believe that AI can bring many efficiencies and bring projects to life that have otherwise been impossible within the confines of the existing labor market.
Undeniably is the role AI will play in changing the way developers work. Tools such as Codeium and GitHub CoPilot show how powerful they can be in streaming software development. Software engineers who embrace this change will be able to create new, groundbreaking experiences. Having more time to envision creative products. However, this can only be established if processes are put in place that can handle the exponentially growing output.
SaaS disruption
Codeium and GitHub operate in the software as a service (SaaS) space, which will see its own disruption. While this is an analysis all on its own, it’s worth mentioning, as many enterprises rely on them today. The industry itself employs countless people around the world. In June 2024, McKinsey predicted that generative AI would have a greater impact on the software industry than the introduction of SaaS. The impact of AI is felt more rapidly than software as a service, seeing a projected penetration rate of approximately 10 percent over the course of just four years, translating to an impact 3 times greater than SaaS did before.
McKinsey notes that software leaders around the world will have to brace themselves for rapid changing economic forces, which result in shifts in user segments, value pools and industry dynamics. The consulting firm already witnessed disruption in the software space, which is replacing a select number of software applications. While the full scope of the disruption is hard to predict, McKinsey adds, generative AI will result in value creation and erosion within the industry.
The consultancy firm predicts that spending on generative AI will increase to approximately $175 billion and $250 billion by 2027. On paper this sounds promising, however, this entails a spending shift, with investments moving away from established players toward new vendors who provide novel technologies. If buyers decide to stay with their current suppliers, they will force them to adopt Gen AI features. If these providers fail, the expected churn at established players will increase by one to three percentage points. Hence, players such as GitHub, are pushing AI solutions to remain relevant to their customer base.
