Published on the 24/09/2025 | Written by Heather Wright
And where does low-code fit in?…
AI-assisted coding is continuing to gain ground, but many IT leaders are concerned that its growing use will stunt skills growth – and businesses shouldn’t rule out low-code either.
A Canva survey of 300 technology decision makers, including 50 in Australia, shows while the vast majority of those surveyed believe AI tools are transforming developer skill sets, 21 percent were concerned that use of the technology may stunt junior engineers’ development.
“All software engineers will need to evolve and become AI-native software engineers.”
Nearly one-third of the IT leaders also cited over-reliance on AI without developer accountability as their top concern, with 95 percent flagging potential risks when deploying AI-generated code without sufficient review.
The survey shows technology leaders are clear on the need for human oversight and guardrails around maintainability, security and quality.
Brendan Humphreys, Canva CTO, says the engineers who will thrive in this new era are adopting AI to enhance their thinking and output, not replace it.
“They’re bring creativity, critical thinking and a deep sense of ownership. That’s how organisations will move fast, innovate and build responsibly.”
He says AI in engineering is a ‘real disruptive change’ that can’t be ignored.
Among those surveyed, AI adoption was nearly universal, with 92 percent using AI-assisted coding tools and 78 percent of developers relying on them every day. That’s set to accelerate with 66 percent of respondents expecting their organisation’s use of AI coding to increase ‘significantly’ over the next 12 months.
Software development was one of the first areas to adopt generative AI, with bold claims early on that AI would reduce demand for human engineers or supplant them entirely. Productivity gains are a key focus for many – in the Canva survey 64 percent cited that as a primary driver for AI-assisted coding tools.
The report highlights the benefits of AI-assisted coding, including rapid prototyping, quicker iteration, the ability to explore new ideas and generate first drafts of code.
The company has previously noted that nearly half of its frontend and backend engineers are daily users of an AI assisted coding tool.
But other recent reports highlight that AI in software development, like in so many other areas, is currently failing to live up to the hype, with Bain & Company’s newly launched Technology Report 2025 reporting productivity savings have been ‘unremarkable’, with teams reporting a boost of just 10-15 percent.
It says early initiatives have often fixated on code generation – using genAI to write code faster – but notes that writing and testing accounts for just 25-35 percent of time from initial idea to product launch.
“Speeding up these steps does little to reduce time to market if others remain bottlenecked,” the report notes.
A randomised controlled trial by non-profit research group Model Evaluation and Threat Research earlier this year found that actually, AI made developers slower – taking 19 percent longer when using AI tools than without, though the developers themselves believed AI had sped them up by 20 percent.
It suggests time spent checking and correcting errors voided any time saving.
Adrian Leow, Gartner applications and software engineering leaders group vice president, says AI is augmenting, not replacing, developers.
“All the software engineers in the traditional sense will need to evolve and become AI-native software engineers,” he told iStart.
“Code assistants require human intervention,” he adds, noting that agents are autonomous or semi-autonomous.
Is low-code defunct now?
He notes too the continuing role of low-code platforms, and says the most common question clients are asking at the moment is whether AI will make low-code tools obsolete.
“Our straight out answer is no. They won’t replace low-code tools.”
He says low-code platforms offer a lot of structured governance tooling, and other tools which come with low-code, that can act as a ‘scaffold’ around what you can do with AI code assistant code generation tools, LLMs that are being used for code generation.
“Hallucinations, bias and all that exist. The different governance tooling and security that come with low-code can help provide more guardrails around how that code is actually developed in that manner.”
He notes that all the big players, including OutSystems, Mendix, Microsoft and ServiceNow, have different agentic AI either integrated, or can have third party integrations to agentic platforms.
“If you are using low-code you will be able to have AI code generation tools and agentic AI tools as well. It’s not like one of the other.”
Canva’s report notes organisations are tightening oversight with human-in-the-loop deemed critical.
Ninety-three percent say AI-generated code is always or often reviewed before merging, and more than half (58 percent) have placed responsibility for AI coding governance at the CTO or CIO level.
Canva says the findings underscore that AI should amplify, not replace, human ability.
“The greatest value emerges when AI tools are paired with human judgement, imagination and expertise.”
