An AI startup CEO says she no longer has engineers at her company. Instead, they have become engineering managers who manage AI agents.
Loading audio narration…
Tatyana Mamut, CEO and cofounder of AI agent startup Wayfound.ai, says her team of two engineers is more productive than a large engineering team. Mamut, a former Amazon Web Services director who managed engineering teams of over 30 people there, says her two engineers can “ship code much faster.”
“The two engineers ship more features than my team at Amazon in 2017,” Mamut said.
Now, coding agents like Claude Code do the engineers’ work, while humans manage them. The team of two, including the CTO, also spends time talking with customers to build custom features.
Mamut chatted with Business Insider about how the startup made this happen and what Wayfound gained.
Wayfound turned its engineers into managers
Wayfound started this transition in 2024. Its engineers initially used ChatGPT to help them code. When other coding tools like Claude Code, Vercel, and Cursor rolled out, the engineers started testing them. Now, they mainly use Claude Code.
“At a certain point a few months ago, Claude Code just became really good,” Mamut said.
She says the engineering team achieved significant efficiency. They don’t have to go through “weeks and weeks of testing” — now coding agents do that. They also use Wayfound’s internal agents to monitor code quality and suggest improvements for engineers.
When Mamut previously managed large teams, the process took weeks. Engineering managers plan project requirements and priorities, allocate work, manage quality, and collaborate with product managers. In the process, engineers may even get into arguments.
“Now we go into a meeting on Mondays and Wednesdays, and we talk about what are the things we need to ship,” Mamut said. “What are the customer priorities? That takes 20-30 minutes. They prioritize what we need to do. Engineers take it and go to Claude Code and say, ‘Here’s what we need to do, here’s what we need to accomplish.'”
Coding agents also helped the startup save money. Wayfound has not had to hire additional customer success employees, as engineers interact with customers.
“Engineers now do a lot more conversations with customers and a lot less staring at screens,” Mamut said.
Mamut describes these employees’ roles less as engineers and more as a “builder” — a title becoming more common at companies like Meta.
“The functions of product, design, and engineering are collapsing into one function,” Mamut said. “The product managers talk to the customers. Designers design out the features. Engineers code out the features. Now one person can do all those things.”
Mamut warned that “agent slop” can become a major problem if organizations use AI tools and agents without constant training and supervision by both humans and specialized AI agents. She said that if companies deploy agents and then walk away, it can “create massive problems.”
“Though it’s easier than ever to start prototyping new AI agents across organizations, very few will actually perform and drive ROI without consistent supervision and improvement,” Mamut said.
Agents will be the future of SaaS
Mamut also had plenty of thoughts about the traditional software-as-a-service business model. Companies like Salesforce, Atlassian, and Workday saw their stocks take a hit as investors worry that AI tools disrupt, or even replace, software.
Mamut said the SaaS model must completely change. To adapt, she says these companies should become “agentic.”
“Board members are requiring that SaaS companies cut costs in order to invest in AI,” Mamut said. “The only way they can cut costs is head count. That’s why we’re seeing a dramatic decrease in head count and in hiring.”
The companies that win will add AI agents to their products, and if they try to hold on to the core SaaS model, they “will be dead in five years,” Mamut said. That means they should invest money, time, and AI tokens to build these agents. They may take a hit on their margins now, though it will pay off later.
Since companies seek to cut costs and head count, they don’t want to be locked into multi-year contracts — a key part of the traditional SaaS model — because they don’t even know how many seats they’ll need down the line.
“That’s a shift that most SaaS companies need to grapple with,” Mamut said.
While AI agents can perform some tasks that humans have traditionally done, Mamut sees humans as managers of agents, builders of customer relationships, and people who set company strategy.
“People want to buy from other people,” Mamut said. “People want to know who is a person they can trust. They still need to build those relationships.”
Have a tip? Contact this reporter via email at [email protected] or Signal at rosal.13. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.
