Artificial intelligence and globalization are shaking up the world of software developers

These changes are important because software talent is highly valued. Salaries are high (see Chart 2).

The average salary for a developer in America is in the top 5% of all professions; This means coders earn more than nuclear engineers. Tech giants need these to make their platforms more attractive; Non-tech company bosses increasingly want coders to help with digitization efforts that they hope will increase productivity and appeal to consumers. Therefore, two shifts is welcome news. The future looks like there will be more and more productive coders and cheaper software.

New technologies have often helped developers. For example, the internet has put an end to the time-consuming task of answering questions using textbooks. Generative AI still looks like a bigger step forward. One reason it is particularly useful for developers is the availability of data. Online forums like Stack Overflow have a huge archive of questions asked and answered by coders. Answers are frequently rated, which helps AI models learn what is useful and what is not. Coding is also full of feedback loops and tests that check whether the software is working properly, notes Nathan Benaich of Air Street Capital, a venture capital (VC) firm. AI models can use this feedback to learn and improve.

The result has been an explosion of new tools to assist programmers. PitchBook, a data provider, tracks about 250 startups producing them. Big tech is leading the charge. In June 2022, Microsoft-owned GitHub launched Copilot. Like many tools, it can deploy lines of code at will. Nearly 2 million people pay for subscriptions, including employees at 90% of Fortune 100 companies. In 2023, Alphabet (parent company of Google) and Meta (parent company of Facebook) launched competitors. This year, Amazon and Apple followed suit. Many companies have developed artificial intelligence coding tools for internal use as well.

However, the usefulness of artificial intelligence is still somewhat limited. When Evans Data, a research firm, asked coders how much time technology had saved them, the most popular answer given by 35% of respondents was between 10% and 20%. Some of this is due to mass production of simple “common” code, but the tools are not perfect. A study by GitClear, a software company, found that code quality has declined over the past year. Use of AI is suspected A survey by cybersecurity firm Synk found that more than half of organizations have discovered security issues related to weak code generated by AI, and AI is still unable to solve tougher programming problems.

New generation vehicles should be better. In June, Anthropic, an artificial intelligence startup, launched its newest model, the Claude 3.5 Sonnet, which is better than previous models at coding, among other features. ChatgGPT’s maker, OpenAI, released a version of its latest model, o1, on September 12, claiming it “excelles at accurately rendering and debugging complex code.”

AI tools can increasingly help with other mundane tasks, like writing notes about what the code does or designing tests to make sure the code won’t fail. Writing code is just part of the job. A software engineer, which accounts for about 40% of their time, according to consulting firm Bain. These tools can also help programmers become more flexible by switching between coding languages ​​more quickly, allowing them to more easily apply their skills to different situations. It says it has seen engineers switch from one language to another in a week instead of three months, reports Prosus, an investment firm. Amazon recently said it saved $260 million when it converted thousands of applications from one type of code to another. AI.

The newfound flexibility extends to different types of programming. Previously, a small application might require a team of six people to work on different parts of the program, such as the user interface or the plumbing of the software. Jennifer Li of Andreessen Horowitz, a major VC firm, says she’s seeing more startups being built with fewer people because programmers can more easily handle many different tasks. Many IT executives say training newly hired developers on the features of their company’s software is also getting faster.

Most of these appear to provide support to inexperienced engineers (see Table 3).

They will be able to perform more complex tasks faster, and some of the work they used to do can be done by laypeople. The growing trend towards “low-code, no-code” platforms that allow anyone to write software will also be powered by AI. Brazilian lender Banco do Brasil uses such a system to allow employees to develop hundreds of pieces of software. Like apps that make it easier to help customers searching for insurance products.

Another consequence of the coding revolution is that young developers in rich countries will face fiercer competition from abroad. According to Evans Data, the number of computer programmers in the Asia-Pacific region and Latin America is expected to increase by 21% and 17% respectively between 2023 and 2029, compared to 13% in North America and 9% in Europe. The imbalance means the rise in offshoring and outsourcing is likely to continue. Consulting firm Everest estimates that about half of all its expenses, including much software development, are outsourced. Other firms that maintain these services in-house have instead established their own outposts abroad to take advantage of lower wage costs. India is the power center of the world. In 2023, exports of software and related services reached $193 billion, half of which was to America.

This helps companies control costs. “This is a very good way to scale without blowing up budgets,” says Shashi Menon, who heads digital efforts at oil and gas services company Schlumberger. About half of the engineering team is based in Beijing and Pune, India.

Offshore capabilities are becoming increasingly sophisticated. Some foreign outposts now provide high-end fares as well as basic software. Sanjeev Jain of Wipro, an Indian firm, says his engineers helped build Microsoft’s video streaming service Teams, as well as designing chips and software for “connected cars” that communicate with other services and devices. software; ai nous itself is something they can sell. Another Indian firm, Infosys, said it recently won a $2 billion, five-year contract to provide artificial intelligence and automation services to an unnamed client.

It’s still unclear what all this means for developers. One vision is for AI and offshoring to take jobs away from Western software developers en masse. This seems very difficult. A huge amount of technical know-how is still required to put together pieces of code and check if it works.

A more optimistic view is that the most tedious parts of making software are done by computers, while the developer’s time is spent on more complex and valuable problems. This may be closer to the truth. Meanwhile, the trends are welcomed for customers. IT executives have long said their bosses want more digitalization with increasingly tight budgets. Thanks to artificial intelligence and offshoring, this may not be too much of a question to ask anymore.

© 2024, Ekonomist Newspaper Ltd. All rights reserved. Published under license from The Economist. Original content can be found at www.economist.com.