DEVOPSdigest

Contrary to what you've read, generative AI (GenAI) doesn't spell the end of software developers.

Yes, IT has joined a list of business operations earmarked for AI rollout as executives buy into the idea that well-trained algorithms can make their operations more "efficient."

Rather than eliminating software engineering, AI is transforming the role, and with employers expecting to increase hiring, the question is how developers leverage AI and adjust to the new role.

Ready Or Not, Here AI Comes

AI has come of age after years in the business-technology wilderness thanks to the success of GenAI.

Two-thirds of organizations claim to be using GenAI in more than one business unit — an increase of nearly a fifth in the last year. Software development is in the vanguard of a group of technology functions considered ripe for AI. The number of programmers using AI has hit a record 62% — up from just fewer than half a year ago.

Many are using assistants in core areas: to generate code snippets and documentation, for refactoring and code standardization. Find faults and vulnerabilities, to improve security and stability. And analysis — to predict problems before code hits production environments.

As a result, AI assistants are redefining the developers' role in two ways: first, by taking on the repetitive aspects of the role that are important yet tedious and time-consuming. Second, by giving developers a new set of responsibilities: dealing with assistants' shortcomings. Agents struggle to work creatively in interpreting context to build systems, they generate false positives in testing, and they can produce code that doesn't make sense — better known as hallucinations. To deal with this, developers are having to become more like supervisors — reviewing output and maintaining a watching brief on the effectiveness of assistants.

Warning: Improvements Ahead

In the long term, assistants will improve and AI will become part of the technology infrastructure relied on by digital businesses. This will create demand for those with the skills to develop and maintain that infrastructure. Already, 82% of IT leaders say increased recruitment for those with AI skills while four in ten expect to reskill at least 20% of their workforce, according to McKinsey. The best way for developers to adapt is by taking on new, and adapting a clutch of existing, skills in a number of key technology and non-technology areas.

On the technology side, it will be important to master two new skills particularly relevant to AI. The first is prompt engineering. By crafting and fine-tuning effective prompts engineers can go a long way to ensuring their systems produce high-quality, relevant results. Excellence in prompt engineering will also speed up review and editing, getting AI systems into production faster.

Second is a foundational understanding of large language models (LLMs) and small language models (SLMs). Much of what teams need to know about today's LLMs isn't new and should be expected to change. SLMs, meanwhile, are a new field with models changing fast. Due to the low computational cost and faster inference times, SLMs enable decentralized AI architectures, where multiple AI agents can run independently. Agentic workflows enhance efficiency, scalability, and decision-making by automating tasks, optimizing resources, and enabling intelligent human-AI collaboration with adaptability and precision. It will therefore be important that engineers know models' latest features and capabilities to build, train and tune them.

Design of data structures, algorithms and systems - existing developers' skills - will take on a new significance. Organizations will expect their AI products to serve customers and beat the competition. This will take efficient and scalable data structures and algorithms to handle large amounts of data and process computational tasks quickly. System designs will be expected to meet customer and business needs, be secure and simple to maintain. This will put engineers center stage.

Going hand in hand with this will be an elevated form of code review. As the pressure increases to make differentiated AI products and services, those reviewing code will need to measure them accordingly. It will be necessary to assess systems against criteria like business objectives and architecture. Expect this role to fall to those with a knowledge of their organization and sector they serve.

Underpinning each of these is the need for a raft of non-technology skills. These are areas engineers can grow into to support the AI infrastructure. Activities like shaping and tuning language models, creating streamlined data structures, and reviewing code will mean developing an understanding of product design and of business strategy. It will mean expanded critical analysis to detect the first, early signs of bias and hallucination before they become dangerous. And it will mean improved powers of communication and collaboration to work with mixed technology and business teams in building, refining and maintaining AI that's fit for business.

Conclusion

AI is recalibrating the role of software engineers — taking away some functions and bringing new responsibilities. As AI turns into an enterprise platform, it'll be the engineers working at this evolved level who'll be expected to ensure AI systems deliver the outcomes business expects.