Demand for engineering managers is surging in the agentic coding era

Key takeaways:

• Engineering managers are becoming elite Individual Contributors (ICs) because their skills in delegation, context-setting, and task breakdown map perfectly to overseeing AI agents.
• Engineers now need tech lead skills to manage parallel agents, leading to a collapse of traditional hierarchies.
• Firms are reducing management layers, expecting engineers to act as “mini CEOs” who own end-to-end product outcomes.

The world’s software engineers have all been promoted in one fell swoop. Suddenly, with the rise of agentic coding, everyone’s now a manager. There’s just one difference.

“Instead of having teams of humans working together, each individual contributor ends up being like a manager. But instead of a manager of humans, they’re a manager of agents,” said Scott Breitenother, co-founder and CEO of agentic coding startup Kilo.

As agentic coding takes off – wherein engineers describe what they’re trying to achieve to AI agents and send them off to write, test, and debug code for them end-to-end – engineering managers are finding that this new approach resembles their work far more than it does the traditional developer role of sitting at a keyboard and writing code by hand.

Because of how neatly their skillset of breaking down engineering processes and delegating to others maps to agentic coding, some feel this is giving engineering managers and tech leads a distinct head start in today’s talent market. It’s also putting them in high-demand to return to acting as individual contributors, as fast-moving startups in particular look to build teams well-versed in this kind of workflow.  

Engineering leads working to help their teams adapt to agentic coding are teaching management skills to their ICs, and in some cases, are also finding their help is needed less than ever once their teams get the hang of it. Taken together, it all points to the early days of a significant shakeup. Engineering roles, responsibilities, and career trajectories are being flipped and collapsed.  

“Now is the time to let go of role definitions and stereotypes,” said Vinay Perneti, VP of engineering at Augment Code. “Instead, I think you really want to embrace outcomes.” 

Think like a manager

In recent breakout meetings, Perneti heard engineers on his team express that often when they’re trying to code with agents, “the model just doesn’t get it.” So he asked them: is it that the model isn’t capable, or did it not have enough context?

“The observation coming out of that was that the humans who are working with agents, basically, they need the tech lead skills,” he said.

Everyone who has become a manager or tech lead has had to learn how to delegate tasks, including how to ensure someone has both the skills and context required to successfully execute, he explained.

“As tech leads, you have to fine tune this person for this task. What is the right level of delegation? What is the right way to engage with this person so that they have increased confidence? I think agents have a very similar thing,” Perneti said, adding that he’s noticed that people who have been in tech lead roles before are picking up agentic coding much more naturally than those who haven’t. 

Vernon Richards, a senior engineering manager of QA at Phrase, has also noticed the striking resemblance between his role and what he needs the engineers who report to him to do with agents. He’s positioned the shift to his team by telling them they’re now all managers to “robot” teams.

Overall, he said agentic coding requires them to kick the habit of “doing” and instead default to delegating, task analysis, and figuring out how to tease out and clearly communicate the goal. This includes thinking through and articulating the answers to questions like: what tools do I need to give the agents? How can I describe what good looks like? How do I set this up for the outcome I want?

“I’ve got to get them to extract it out of their heads,” said Richards. 

This doesn’t mean engineering managers don’t have anything to learn. There’s already a larger shift starting to unfold as the technology progresses from more task-based agents to always-on agents like OpenClaw. Even the most seasoned engineering and technical leaders haven’t managed that kind of entity before, pointed out Perneti. But as many engineers struggle to adapt to this new way of working, their managers are already halfway there.

From manager to IC?

One engineering manager at a fintech startup, who asked to remain anonymous, echoed the idea that those in his role have an advantage in adapting to agentic coding. He’s used to guiding others and multi-tasking. Kicking something off and then returning to it when it pings you is very natural to him, he said. At the same time, this shift is shaking up his work in a totally different way: he has less of it to do than ever before. 

Before AI, a big part of his job was routing, but he said AI has largely solved that. For example, an engineer on his team was recently assigned a ticket in a domain they’d never worked in. Previously, he’d have to spend between thirty minutes and an hour talking the engineer through which systems are involved and other information they’d need to tackle it. This time, however, the engineer was able to simply show the ticket to an agent, which then compiled everything they needed to know.  

“I’m managing more people, but the amount of management I’m doing per person is much lower,” he said. 

The good news for engineering managers who find their place in the engineering org may be shrinking is that they’re in high-demand as ICs. At Kilo, which launched in early 2025, for example, every new hire brought on as an engineer was previously an engineering manager or senior peer mentor. And that’s by design; looking at how their background mirrors the skillset of agent-first work, Breitenother said they wanted to hire managers only. 

“Yes, they had to adjust to this agentic world where you’re managing agents, but it wasn’t as big of an adjustment,” he said. “I think it’s easier to go from managing a team of humans to managing a team of agents, than to start from managing no one.”

For the same reason, Perneti similarly said he’s focused on hiring people with tech leadership experience. He’s also prioritizing recent graduates. 

“There are two cohorts of folks that are very well suited to take on the AI wave. One of them is the folks who have tech lead experience and know how to have multiple agents running in parallel,” he said. “[The other] is the new college grad cohort that is learning to only build with agents. The creativity that I’m seeing internally from that group is off the charts.”

The great engineering org collapse 

Several leading tech companies have been significantly thinning out managers. Google recently cut 35% of managers who oversee fewer than three people, and Amazon mandated that every team must increase its ratio of ICs to managers by 15% by Q1 2025 – and the firm has continued to lay off managers of all types since.

At the same time, staff and principal engineers at top companies now regularly out-earn engineering managers, sometimes by 15 to 25%, showing increasing demand for top ICs, according to IT hiring firm HR Oasis.

If ICs are now managing and managers are becoming more useful as ICs, what does this all mean for the future of engineering roles and career ladders?

The fintech engineering manager believes there will still be engineering managers for larger teams (because someone needs to do the “people stuff”), but he can’t imagine the role will continue to exist for small teams with just a handful of engineers for much longer. In addition to becoming ICs, they can also transition into becoming project managers, he said. 

Interestingly enough, others see the project manager role as another that will fade away – or more specifically, converge with the engineer role. When Breitenother and his co-founder, former CEO of GitLab Sid Sijbrandij, launched Kilo last year, they specifically made the decision to skip the product management layer entirely and instead have their engineers operate like “mini CEOs” of their own products, including taking on product discovery, prioritization, and other work that was traditionally held by product managers.

It’s a “bizarre, giant transition” for people who have been in the industry for a long time, said Breitenother, but “I think it’s a growth opportunity for everybody.” 

Overall, Perneti believes these changes are collapsing software roles into one role. Engineers and managers – or whatever new role comes from the merging of it all – will need to own end-to-end outcomes, and they’ll be more connected than ever to product strategy and aligning with the company’s overall direction. 

“I think it’s very liberating, if you ask me. It no longer constrains you into a small box,” Perneti said. “For the right mindset, I think this can be very liberating.”

New York • September 15-16, 2026

Speakers Camille Fournier, Gergely Orosz and Will Larson confirmed 🙌

Explore