Software developers say AI technology helps them work faster—sometimes

Artificial intelligence’s ability to write computer code is already having a direct impact on the tech industry—though maybe not in the ways you might assume.

A pervasive fear about generative AI in general is that it might take over human tasks, causing widespread job displacement. But local tech firms that have started to use the technology offer a more nuanced take on its impact on their industry.

The technology can be a time-saver, they say, but is more useful in some cases than others. And, with the caveat that it’s difficult to make long-term predictions about ever-changing technology, the firms also say they view generative AI as a tool rather than a job killer.

“Everyone always asks, ‘Is the AI going to take my job?’” said Wes Winham, founder and CEO of Indianapolis-based human resources tech firm Woven Teams. “And I think the answer is, ‘The AI is not going to take your job. But someone who can use the AI better than you might take your job.’ Folks who don’t learn how to use these new tools, they’re definitely missing out.”

Woven Teams has developed a skills-testing platform that clients can use to streamline the process of hiring software engineers. The client sends its applicants to Woven’s website to complete required engineering skills tests, which frees the client from having to conduct that testing in-house.

Currently, Winham said, only about 5% of Woven Teams’ clients have added AI skills to their requirements for new hires. But he predicted that, at some point, AI will become a required skill for software engineers.

Winham also said companies might be reluctant to add too many requirements to their job openings because they don’t want to discourage potential candidates. Despite the ongoing downturn in the tech sector, he said, good software engineers remain in high demand. “It’s still a very competitive market.”

Winham said his company has also been “experimenting rapidly” with generative AI tools like OpenAI’s GPT-4 and GitHub’s Copilot Chat as aids for the company’s own engineers and has seen dramatic productivity gains from them.

‘Fundamentally different’ tool

Indianapolis-based tech firm Encamp Inc. has also embraced generative AI. The company offers online tools to help customers with their environmental compliance requirements.

Josh Moyer, a principal software engineer at Encamp who works remotely from suburban San Francisco, said the company’s engineers were skeptical about AI’s usefulness until GPT-4 was released in March.

“When [GPT-4] came out, it took us a few weeks to understand that something was fundamentally different about it—that it could essentially reason,” Moyer said. “And a lot of our day-to-day work as software developers is about reasoning through relatively complicated problems and coming up with relatively novel solutions. So we thought, ‘There’s no way artificial intelligence is going to help with that.’ But it does.”

Now, Moyer said, most of Encamp’s software developers make daily use of GPT-4 or a tool from AI company Anthropic, called Claude. “It’s the kind of thing where, honestly, once you get it in your hands as a software developer, you’re like, ‘Please never take this away from me.’”

Moyer said he’s found the tools to be most useful in completing mundane tasks, or for helping him get up to speed on an unfamiliar topic. For some tasks, he said, the AI allows him to work up to 10 times faster—though that varies greatly depending on the task.

Fishers-based tech firm Six Feet Up Inc. has also seen good results incorporating generative AI into its software development process, though this wasn’t the case initially.

When Six Feet Up started using generative AI tools about eight months ago, the company found the technology’s usefulness to be “marginal at best,” said Chief Technology Officer and co-founder Calvin Hendryx-Parker.

Generative AI has improved markedly over the months, Hendryx-Parker said, and now the company’s developers are gaining a 20% to 25% boost in productivity by using tools like OpenAI’s ChatGPT and GitHub’s Copilot Chat.

Six Feet Up develops software and applications for clients in a range of industries, including life sciences, education, government and others. Founded in 1999, the remote-only company has about 25 employees, seven of whom live in Indiana.

Hendryx-Parker said Six Feet Up has found generative AI to be helpful not only in writing code but also in testing the code and documenting it. In software development, documentation includes words or illustrations that explain the code, making it easier for developers to understand the code when they come back to it later.

Mixed results so far

Meanwhile, Carmel-based software development firm Software Engineering Professionals Inc. has experimented with generative AI but isn’t yet using it to build projects for clients.

“These tools, they are useful, but they don’t match the hype that is being attached to them,” said Aaron Alexander, delivery lead at SEP. “For us, we saw small enough efficiency gains that it didn’t really seem worth pursuing.”

Alexander said he’s not sure why SEP saw relatively minor productivity gains from AI, while other companies have seen dramatic improvements. He said AI might be more impactful for small firms than for a company the size of SEP, which has 160 employees, most of whom are directly involved in software development.

It also might be the case that generative AI is less helpful in creating large and complex projects, Alexander said, “but I don’t know if I feel super confident in that thesis.”

Whatever the reasons, he said, it’s likely that those efficiency gains will improve as the technology evolves.

A recently released study from New York-based consulting firm McKinsey & Co. backs up what local firms have discovered—when it comes to software development, generative AI is more useful in some situations than others.

The McKinsey study, released in late June, was based on research with more than 40 McKinsey software developers with varying degrees of experience. The developers were asked to complete different software development tasks, both with and without the use of generative AI.

When using generative AI, the developers were able to generate code 35% to 40% faster, and they were able to document code 45% to 50% faster. When it came to code refactoring, or improving existing code, the developers were 20% to 30% faster when they used generative AI.

But for highly complex tasks, generative AI improved the developers’ efficiency less than 10%. And in some cases, the study found, developers with less than a year of experience found it took them longer to complete the tasks when using generative AI.

New ways of working

One thing the firms do agree on is that the use of generative AI changes the way developers do their jobs.

The better the query, the higher the likelihood that the generative AI will produce a good result, Winham said. For this reason, knowing the best way to query generative AI—a skill known as prompt engineering—is an important part of successfully using the technology, he said. “It’s really about being an effective partner with these tools.”

Winham said he’s also found that explicitly asking the AI to analyze its own answer can improve the chance of getting a good result. The developer might, for instance, write a prompt that includes a phrase like, “Think step by step, and check your answer at each step.”

A developer using generative AI will also likely spend less time producing code and more time analyzing it.

One of AI’s current shortcomings includes its tendency to “hallucinate,” producing answers that at times are fabricated or incorrect. This also happens when AI is asked to produce code, Alexander said.

“You really shouldn’t be blindly trusting the code these models generate,” he said. “You’re changing the way you work and needing to evaluate, ‘Did the code coming out of this AI model actually solve the problem? Are there bugs in it? Is it OK to use?’”

At the moment, at least, none of the developers said they fear these new tools will make them obsolete.

Hendryx-Parker, in fact, has the opposite view—he predicted that the overall advance of technology and the new possibilities that generative AI can unlock will actually increase the demand for human tech talent. “Now, maybe I’m just a super-optimist. But I do think there’s just so much more for humankind to go out there and accomplish and conquer, and lots of big problems to still solve.”

Jake Miller, CEO of Indianapolis-based Engineered Innovation Group, said he’s optimistic about generative AI’s potential. The firm, which offers software development and innovation consulting, started using the technology about a month ago.

Miller said he sees generative AI as something that will make programming more accessible to more people. He also envisions that the experience of software development will be radically different in the next 10 years or so. Perhaps rather than using a keyboard and a mouse, people will verbally tell a computer to solve a problem and get an answer without having to write code.

What place would Miller have in a scenario like this?

“I don’t know the answer to that,” he said. “But I hope it’s something fun.”•