This post was originally published on this site.
Does This Signal A Workforce Revolution?
The swift progress of Large Language Models has significantly altered the software engineering field since ChatGPT’s burst in late 2022. In 2023, I moderated a panel featuring Maxim Fateev, co-founder and CEO of Temporal IO, an open source ‘durable’ execution system for application development and Anand Kulkarni, CEO of Crowdbotics, a platform for secure app development using systematic code reuse. Both shared their views on the impact of AI on the role software engineers.
One year later, we reconcile their perspectives as LLMs and AI-assisted development tools continue to reshape how software is created. We spoke with several experts to understand these changes: including Paulo Rosado, founder of OutSystems; Andre Shojaie, Founder and CEO of HumanLearn; Qaiser Habib, Head of Canada Engineering at Snowflake, and Dr. Christine Colclough, founder of The Why Not Lab, a political economist and expert in AI’s impact on labor.
Have Kulkarni’s and Fateev’s perspectives hold true today? We examine how Generative AI has evolved one year later as we dive into the software development landscape, AI integration, organizational change and the broader effects of technological progress. And, from a 10,000 foot view, what is this signalling for what’s to come for jobs across sectors?
The Changing Nature Of Software Development
Maxim Fateev, CEO of Temporal.io, in 2023, saw the engineer’s role getting harder. He explained that as technology evolved, they spent a lot of time trying to improve deployment in operations, observing “Today, systems can deploy 10 times the volume of processes at higher rates, But the core business abstractions didn’t change. We began building more distributed services. So, every developer right now is distributed system engineer… And it means that the life of developers got harder because you don’t have transactions anymore. You need to call all these APIs. Consistency is a problem.”
In 2023, Kulkarni emphasized a shift in the engineer’s role, acknowledging the sea change that is forcing engineer to think about where their role is in the software development process. He reflects, “Before you might have just needed to be an architect or a creator involved in developing individual lines of code and individual functions. Now, because tools are able to assist in letting us create more and more rapidly, we must become editors as software engineers. We must become system level thinkers, as opposed to the individual… [where] hands-on keyboard is the most important [role]. And that’s a surprising shift.”
Present day, Andre Shojaie, Founder and CEO of HumanLearn, with expertise in AI, and organizational transformation, echoes this sentiment, noting: “Software engineers, once viewed primarily as coders, are now expected to operate at a higher level of abstraction, requiring a blend of creativity, critical thinking, and problem-solving that AI alone cannot replicate.”
AI Tools Are Changing The Game
In 2023, Fateev and Kulkarni acknowledged the growing importance of AI-powered tools like GitHub Copilot. Fateev predicted: “AI will be huge helper, Copilot will be absolutely there and I’m pretty sure Copilot will be much more useful. But that engineer will be assembling those solutions… on a higher level of abstraction.”
Qaiser Habib, Head of Canada Engineering at Snowflake, concurs and shares: “As a point of reference, Snowflake’s CIO Sunny Bedi asked his development team to estimate how much of the code they are currently writing could be handled by an AI tool, and the consistent answer was 30 per cent. This estimate scratches the surface if we consider how AI-generated code can also be leveraged for reusability, sharing, testing and quality assurance.”
Shojaie observes today, “Tools like GitHub Copilot or AI-based code completion and review systems are becoming commonplace, aiding developers but also shifting the skill sets required. As roles evolve, engineers and technologists are finding that their responsibilities are becoming more interdisciplinary, necessitating knowledge in AI ethics, system design, and even business strategy.” This broadening of scope, as Shojaie indicates, ensures the engineer’s expertise remains relevant as AI becomes more capable of handling routine tasks.
Increasing Relevance of Low-Code, No-Code AI Integration
In 2023, Kulkarni was skeptical about the impact of low-code and no-code solutions in producing complex software systems, expressing: “The promise of low code, no code was great. We had this idea that you could have people, who are not technical, build powerful software systems and apps. And all engineers know this is sort of a myth. It’s proven that we can build out simple things using low code. But no great products have been built on low-code, no-code systems. Ever.”
However, the integration of AI with low-code platforms in development of enterprise systems has broken new ground. Paulo Rosado, founder of OutSystems, and a pioneer in low-code software development argues: “We’ve been compressing projects from four years into seven months.” He elaborates on the impact of AI integration: “We’ve been incorporating AI, especially Generative AI, to enhance our products. Our platform now includes various AI agents and generic AI technologies. What’s particularly interesting is how business applications are changing with the AI agents in the middle. Traditionally, digital systems were built with components like portals, workflows, logic engines, business rules, and data repositories. Now, we’re exploring where to integrate AI agents into this architecture to transform how digital systems are constructed.”
Rosado adds that building a new AI agent is merely the initial step. To be effective, the agent must be surrounded by supporting logic and software that can be readily integrated into existing business processes. His team observed that a considerable amount of work is required to create a system that is both usable and adoptable. The result is often a hybrid solution that, for example, combines a portal interface layered on top of an AI agent, or a set of policy rules governing the agent’s data access.
Adaptation Overshadows Job Vulnerability
Habib, of Snowflake, does not worry that automation will replace the work of those who develop these solutions, explaining, “AI will augment their productivity, making way for more engaging, high-value work like creating architectures and designs, resolving ambiguity around hard business problems.” He suggests that data scientists are now equipped to create full stack applications, and can go from creating the dashboard, to prompting the LLM to extracting the answers to questions faster: “Today’s data scientists have more room for creativity and deeper analysis — and this could make data science an increasingly attractive and exciting field for young talent.”
Dr. Christina Colclough is the founder of The Why Not Lab, an organization which empowers workers and unions globally, advocating for digital rights and fair policies. She has a differing view when it comes to the illusion of productivity: “We’re currently in a very dangerous situation. For a long time, we’ve been told that AI and digital technologies will boost productivity and efficiency. But who has really proven that? From what I observe, many workers are saying the opposite. Their administrative workload has skyrocketed since these systems were introduced. Now, the new narrative is that we can reduce the workforce by letting these systems take over. So, the message has shifted to: ‘Fire people, and let the systems do the work.’ It’s just a new spin on the same old productivity argument.”
The Future Role of Software Engineers: The Harbinger of Things to Come
In 2023, Kulkarni and Fateev stressed the importance of continuous learning. Kulkarni suggests “This is a new way to think about building software, and it’s a new technology to understand and master how it works… to be relevant.”
Shojaie reinforces this point: “Despite the optimism about AI’s potential to reduce the more mundane aspects of their jobs, there is a palpable concern among technology professionals regarding the rapid pace of technological change. This concern centers on the need to constantly update skills and maintain job security in an environment where AI’s capabilities are continually expanding.”
Habib agrees that AI is driving a push to upskill engineers but argues that for those developers who have yet to develop a ‘deep skillset,’ their learning curve just got steeper, adding “AI does a great job writing a bunch of code, but the designs, blueprints, strategy and the brain surgery code still needs engineers. So, nervousness around AI replacing the role of the engineer applies to those who haven’t yet developed a deep skillset.”
Habib claims that embracing AI can give engineers time and bandwidth back, speeding up the learning process to attain deeper skills. He advises, “For engineering leaders, my advice is to facilitate upskilling in the workplace, which could include implementing certification programs, making educational platforms accessible to all, and encouraging professional development initiatives.”
Shojaie offers context to these trends: “Automation, especially AI, has undoubtedly accelerated the transformation of various industries. Over the last few years, roles involving routine tasks have seen significant displacement due to automation. However, AI has also spurred the creation of new jobs, particularly in AI development, data analysis, and AI system maintenance.”
As the industry evolves, the role of software engineers is transforming. Shojaie provides insight into this shift: “The evolving role of the software engineer will likely involve greater focus on oversight and governance, ensuring that AI systems not only operate efficiently but also align with human values. The challenge lies in maintaining a balance where AI augments human capabilities without diminishing the critical role of human judgment in decision-making.”
He elaborates: “As AI permeates more aspects of daily life and industry, engineers will need to evolve into roles that prioritize ethical AI deployment and the integration of AI into complex, multi-faceted systems. This will require a blend of skills from psychology, ethics, and regulatory knowledge, alongside technical expertise, to navigate the intricacies of AI in a way that benefits society as a whole.”
Colclough reminds that science, technology, engineering, and mathematics have long been regarded as essential subjects, with experts asserting that they would ensure job opportunities as technology permeates every aspect of life, however she argues, “…these are also the most automatable skills.”
Uncommon Logic concluded from their from 2023 research on LLM workforce impact, “…models nearly matched human performance on predictive reasoning tests, heightening concerns about the potential impact of AI on the workforce.” And while we’ve recognized the multitude of large language model hiccups from inaccurate and false responses, lack of domain knowledge or context, bias and fairness issues, lack of fine-grained control, the struggle with reasoning and integration across different domains, to infringements to data privacy consent and copyright laws, to name a few, the demonstrated capabilities thus far continue to heighten concerns of job displacement across many sectors. Colclough reiterates the hype around generative AI, “So yes, some would say that’s because we’re in early days. These systems will learn, but I think there’s a lot of hype and it’s having dangerously so real-life impacts on workers.”
This chart below (June 2024) highlights workers (top right quadrant) that are at high risk of AI displacement. Software developers, on the other hand, have been labelled high AI exposure and low automation risk. While Fateev and Kulkarni in 2023 predicted significant changes, Rosado’s 2024 perspective showed that the pace of change exceeded expectations, with development timelines shrinking dramatically. In addition, the shift from hands-on coding to system-level thinking, predicted by Kulkarni in 2023, was confirmed by Rosado in 2024, with his view engineers will take on more strategic roles. Today’s reality belies this study: software developers are not invulnerable to the threat of automation.
Cities With the Most Workers at Risk of AI Job Displacement
Uncommon Logic
Colclough reinforced the Shojaie’s sentiment expressing, “the skills that are truly required are those that computers can’t replicate–complex thinking, governance, emotional intelligence and ethical considerations.” JP Morgan, nonplussed by these limitations forged ahead to roll out ChatGPT AI Assistants across its 60,000 employee base.
Colclough was not surprised by this news, and declared, “It’s a short-term, reactive approach to think, ‘We can save time by using ChatGPT… The question that arises is: what happens to innovation? What happens to skills development? What becomes of the innovative thoughts that emerge from conversations between people? This knee-jerk reaction fails to consider the long-term implications for creativity and progress in the field.”
The Continuous Need For Human Skills
Dr. Christina Colclough emphasizes the importance of what she terms “inclusive governance” in the context of AI and algorithmic systems. She argues for the necessity of human oversight and control, stating, “If the system cannot be explained, then of course there’s no human control. It’s as simple as that.”
Colclough advocates for a collaborative approach to governing these technologies, insisting that “you have to have humans who understand, can explain, can govern these technologies in cooperation with the subjects of the system.” She illustrates:
“If it’s in the workplace, it’s an algorithmic management system. It should be governed with the workers. If this is an AI system which will have massive effects on software engineers, they should be at the table. If this is a public service algorithmic system to match long term unemployed with jobs, representatives of the long term unemployed should be at the table.”
The concluding argument among these professionals within software development is that the role of the software engineer will not die; instead, the growing AI capabilities will create the need for human creativity, strategic thinking and oversight in software development. Shojaie’s view was optimistic, “This shift suggests that AI is less about replacing jobs and more about transforming them, making roles like software engineering more dynamic and impactful.”
For Colclough, the implications for society are more sobering. “The effectiveness of ChatGPT depends entirely on the quality of the prompts it receives. The more knowledgeable you are about a topic, the more specific and effective your prompts will be, leading to better results. While it’s true that you could use ChatGPT to write an essay about Einstein without having any prior knowledge of him, that doesn’t equate to genuine understanding or knowledge.”
It’s coming and this study predicts that by 2025, 50% of digital work will be automated through LLMs, with an estimated 750 million apps using LLMs.
When the role of the makers of technology, themselves become more vulnerable to displacement, the concern is for those, whose work is not tied to innovation, how much more vulnerable will they be? What will it mean if STEM, which was once the path to future-proof careers, falls prey to what we’re seeing today? We are constantly reminded that we are squarely in a moment of unpredictable change. It is still early day and many kinks have yet to be remedied in a system still evolving. When problems arise, new opportunities are created. This symbolic of the dot-com boom which created new professions and slowly made others irrelevant.
What Colclough reminds us is that we, as humans, need to be accountable for the eventual outcome and reinforcing Shojaie’s call for a blending of interdisciplinary skills, she presses, “We need to break down current silos and across occupations and skill levels, nurture and embrace the diversity of opinion and experience. Rather than a future that subordinates human competencies and purpose, we should all commit to the flourishing of self and others. If we can do this, technological developments will serve humanity rather than destroy it.”