You have 1 article left to read this month before you need to register a free LeadDev.com account.
Estimated reading time: 6 minutes
Key takeaways:
- AI isn’t just cutting jobs – it’s reshaping them. Teams are becoming more fluid, with engineers redeployed to higher-impact, cross-functional work rather than simply reduced.
- Platform engineering is evolving, not dying.
- With AI supercharging output, the real challenge shifts to managing sprawl.
AI-induced layoffs are constantly making the headlines, with large-scale cuts at Amazon, Block, and Meta just a few recent examples.
However, not all changes occurring throughout the software industry are dismal. Some focus more on talent redistribution and reshaping team boundaries.
At least that’s what Jon Thornton, director of engineering at Squarespace, tells LeadDev. He views AI tools as accelerating his group’s agility. “We’ve been adjusting ownership boundaries quite a bit,” Thornton says. “We’ve been spinning up new teams and winding down old ones.”
Thornton directs the platform engineering team, which builds abstractions on top of infrastructure to empower internal product developers. Whereas platform teams have historically built internal platforms from the sidelines, Squarespace platform engineers are becoming more directly involved in migration work and cross-functional efforts with the help of agentic AI tools like Cursor for engineering tasks.
“I’ve seen certain engineers really gravitate toward it,” says Thornton. In one case, he had to free them from existing responsibilities so they could focus on more “leading-edge” work.
In the industry at large, tech analysts have recently suggested that platform engineering has entered a trough of disillusionment, citing low adoption of internal developer platforms and unclear return on investment as hurdles.
At Squarespace, agentic AI tools are altering how platform teams distribute their work, extending the practice beyond the static developer portal and helping teams move faster across the organization.
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
Leading company-wide upgrades
For instance, take large-scale migrations. In the past, deprecating and upgrading platform components at Squarespace was a department-wide goal encouraged by the platform team.
Platform teams would create migration goals, and then wait for developers to comply. However, this created more work for end developers, which in effect constrained the organization, says Thornton.
Now, that process is changing. “Platform teams can do a lot of the work themselves now,” he says, “thanks to the leverage that they’re getting from generative AI-coding tools.”
Squarespace is using Cursor, the AI-coding platform, throughout the company. Now, engineers can upgrade a library simply by using a prompt in Cursor. This lowers the bar to perform upgrades.
The big benefit for Squarespace is being able to perform more migrations, faster, without increasing costs. Thornton says this is helping realize the original goal of platform engineering.
Newfound agility is helping the company maintain and further modularize parts of its microservices stack. However, Thornton notes that it’s still early days, and they’re taking an iterative approach.
“We’re doing it with baby steps, starting with smaller upgrades,” he says. “We’re building confidence that the platform team isn’t biting off more than they can chew when they say they’re going to own 80 or 90% of the migration work.”
Rethinking shared code
Another existential shift for platform engineering is that, using AI for code generation, it’s often easier to vibe-code something from scratch rather than use a pre-existing library or system.
One reason may be that generative tools tend to work more naturally with widely-used external frameworks than highly proprietary internal technologies. Trying to compensate by feeding in a lot of internal context can, at times, confuse Large Language Models (LLMs) and also drain token usage.
“We’re starting to move away from the idea that shared code itself is valuable and saves people time,” says Thornton. Instead, they’re shifting toward prioritizing quality developer experience over implementation details.
For example, they’re exploring alternatives to a shared UI component library in favor of a more AI-driven process. This could help retain visual and interaction consistency while removing some friction when using the original library.
More like this
Slimming the platform engineering surface area
Squarespace is not alone in turning to AI for in-house boilerplate code generation. In early 2026, a Cloudflare engineer used an AI model to reverse-engineer Vue.js, the popular frontend framework, to streamline software builds.
As Cloudflare’s Steve Faulkner writes, “It’s 2026, and the cost of building software has completely changed.”
The barrier to reverse-engineering software components or software-as-a-service has never been lower, prompting engineering leaders to reconsider how much platform surface area they really need.
So, is platform engineering less relevant? “We’re still answering that question,” says Thornton. “The best I can offer now is that we’re thinking about consistency as the real outcome we’re driving towards.”
For him, it seems AI doesn’t make platform engineering obsolete wholesale, but it’s reshaping its initial premise and what a platform actually contains.
“We are actively slimming down the surface area of some of our platforms in recognition that coding tools have a harder time working with our internal microservices library,” Thornton says. Instead, they’re moving toward exposing off-the-shelf components, like Spring Boot, directly.
Culling the software landscape
Now that it’s easier than ever to build and ship new features, more onus is placed on engineering leaders to continually cull their software landscape. This is another area where platform teams are poised to assist.
“If all you do is take this extra velocity that you’re getting from AI tools and ship more stuff, eventually you’re going to end up with an unmanageable pile of features,” says Thornton. AI-enhanced development must be tied to end business outcomes, but it’s a muscle many engineering departments are only just beginning to flex.
One measure of business success is analyzing the percentage of end-user engagement with a feature. At Squarespace, they’ll run tests by temporarily disabling features to see if end users notice, says Thornton. The same tactic could be applied to internal platforms.
“It’s refining the playbook for how you go about deprecating something,” he adds. While in the past it felt like engineering didn’t do a great job of communicating why feature removal mattered, Thornton says the company is getting better at bringing product and business leaders into that conversation.
New York • September 15-16, 2026
Speakers Camille Fournier, Gergely Orosz and Will Larson confirmed 🙌
Cross-pollinating AI knowledge
A strategy that is helping everyone get on the same page is conducting bi-weekly talks on how Squarespace is using AI internally. These are open to both technical and non-technical teams.
Platform engineers are also sharing knowledge of AI best practices and prompt engineering tips within Squarespace, helping individual contributors make the most out of AI tools like Google’s Gemini.
For example, one area they’re experimenting with is having engineers check coding prompts into the repository along with the code they submit as part of their continuous integration pipeline.
“This is helpful from a human perspective, and aids context when you throw an agent at it,” says Thornton.
Another application is documenting step-by-step operational processes, such as how engineers conduct incident postmortems, as a prompt within Gemini. This can help LLMs spot missing context and improve postmortem quality, he says.
“LLMs have been really great at detecting that and pointing it out to postmortem authors,” Thornton adds.
Platform leadership responsibility
“Code used to be expensive,” says Thornton. “It’s becoming a lot cheaper now.” With the dawn of agentic AI, the ability to create code on-demand is causing some to rethink their allegiance to specific stacks. It also means the ‘stickiness’ of vendor ecosystems has shrunk considerably.
AI is empowering individual builders, altering the status quo for software development. One of the more overlooked changes is how it’s shifting team makeups in a headcount-neutral way.
Thornton describes inheriting a platform team made up of two principal engineers and a senior staff engineer. Once they got their hands on agentic tools, they were able to take on more cross-functional scope across other departments.
As an engineering director, you want to delegate tasks to ensure teammates are doing important work within their new capacities. At the same time, you don’t want to overload top performers with more responsibilities.
Agentic tools free up space, but leave an unanswered question: how do you reallocate that time without creating rework fatigue?
Navigating those organizational shifts will likely require ongoing empathy for teammates and continued transparency and communication, especially while high-performing engineers are asked to take on broader scopes. Regardless, platform engineering, like all disciplines, must evolve.
“My intuition for how long or quickly something could be done was starting to feel inaccurate,” Thornton says. After spending more time with Claude Code, he was struck by the possibilities. “You gotta use the new tools,” he says.
