As AI coding agents become more prevalent, companies like Replit, Kilo Code, and Symbotic are working to manage the costs associated with their use. At Kilo Code, engineers are reading or writing code themselves only about 1% of the time now, according to co-founder Emilie Schario — the rest is agents.
Shifting Responsibilities
This shift is forcing new issues onto dev teams: which systems are safe to hand over, who cleans up when models fail, how to support multi-model architectures, and whether high token bills mean real progress or just burned IT budget. Emilie Schario said at VB Transform 2026, “Unless something’s really broken or debugging, 99% of the time engineers are not reading or writing code anymore.”
Jared Go, distinguished engineer for AI and cloud at warehouse automation company Symbotic, noted that the current moment is about directing the focus of AI. He said, “These are my criteria. Let’s look at it from the lens of security, elegance, clean, concise code, water tightness.” That way, AI does most of the heavy lifting, and human code review isn’t as critical.
Human Involvement
Human involvement becomes necessary further down the line, Go noted, because agents don’t make strong product decisions. “Greenfield [building brand new codebases] is easy for agents. Brownfield [writing, updating, or maintaining existing code] is where the actual challenge lies.”
Related: Asana creates secure AI workplace agents
Replit takes a different approach: While the company has “gone very agentic,” they’ve been more conservative with AI coding, explained Amol Jain, head of product engineering. An agent reviews each pull request and assigns it a risk score; low-risk requests are self-merged by their author, while others go to human reviewers who read the code and give feedback.
Multi-Model Support
AI providers are evolving beyond the lock-in model, as customers increasingly demand multi-model choice. Kilo Code supports 500-plus models in its gateway. Schario said, “Your software should be decoupled from the model that you’re using.”
For instance, Schario said companies often use expensive frontier-tier models to architect a project, then switch to a less expensive open-weight model for the rest of the work. It’s also important to respect model provider limitations, such as when they need to work in closed or isolated environments or providers in their specific regions.
As AI adoption increases, runaway costs have become a concern, leading to some enterprises tracking and capping AI use through cost control. Schario said Kilo Code points customers to the same workflow: use expensive models for planning, then open-weight models for affordability.
Cost Management
Symbotic has set per-month cost tiers for its employees. They built a tool that gives managers visibility into requests and usage trends. They can then move users up or down a tier as they see fit, Go explained.
Related: Bright Machines Hybrid Robot Cell Tackles AI Bottleneck
When Cursor — which Symbotic uses heavily — ended a legacy discount that had grandfathered the company into a flat per-request rate even for frontier models, and moved everyone to full pricing, it forced a company-wide reckoning on efficiency, Go said.
Replit broadened agents beyond engineering, and eventually found that a user on the support side had “blown through an insane amount of money,” Jain said. When they looked under the hood, they figured out it was because they were running an automation on GPT 5.5 Pro Max.
They will need to find ways to balance the benefits of AI coding agents with the potential costs.
By implementing cost management strategies and using multi-model support, companies can maximize the value of their AI investments.
