.png){kind=logo}
Anne-Maria Salmela is an early employee at Resolve AI, which raised $35m at seed from Greylock Ventures, and works at the cutting edge of AI innovation. After studying economics and applied mathematics on a full scholarship at New York University, she focuses on how AI is changing how software is built and operated.
As artificial intelligence rapidly reshapes how software is built, Anne-Maria Salmela believes many teams are focusing on the wrong problem.
“AI has made us dramatically better at writing code,” she says. “But that’s no longer where engineering teams are getting stuck.”
Over the past few years, tools for code generation, intelligent IDEs, and automated testing have become ubiquitous. Engineers today are producing far more code than ever before, faster, cheaper, and with smaller teams. Expectations have risen accordingly. Ship faster. Fix faster. Scale endlessly. Do more with fewer people.
Yet while code creation has accelerated, Salmela argues that the rest of the software engineering lifecycle hasn’t kept up.
“The real bottleneck has shifted,” she says. “It’s no longer writing software. It’s running it.”
Despite advances in AI-assisted development, most production systems are still operated using workflows that haven’t meaningfully changed in years.
Teams rely on manual dashboards, fragmented tooling, and the tribal knowledge of a handful of senior engineers. When something breaks, debugging often becomes a high-pressure scavenger hunt - logs in one place, metrics in another, decisions trapped in someone’s memory.
“These workflows drain time, energy, and focus at exactly the wrong moment,” Salmela says. “During on-call. During incidents. When teams need clarity most.”
The result is a widening mismatch: engineers can now write and deploy code at unprecedented speed, but struggle to understand, troubleshoot, and operate that code once it’s live.
“That gap is choking engineering velocity at scale,” she says.
Salmela believes the next major wave of AI impact won’t come from writing even more code, but from transforming how systems are operated.
“AI should be helping us reason about production systems, not just generate new ones,” she says. “Otherwise, we’re just accelerating complexity.”
As software becomes more dynamic and distributed, manual approaches to observability and incident response don’t scale. AI-native operations, systems that can explain what’s happening, surface root causes, and guide response in real time, are becoming essential.
In her view, the teams that win won’t be the ones that write the most code, but the ones that understand their systems best under pressure.
For students and early-career professionals, Salmela says the key isn’t to chase every new AI tool, but to understand where leverage is actually moving.
“Everyone is learning how to prompt,” she says. “Fewer people can still reason and apply critical thinking on a systems level.”
She encourages young people to:
Utilize AI as much as possible but build the muscle to understand the “why” and “how” in all of the outcomes AI generates for you
Question outcomes, and don’t take it for face value
When considering where to work, think about which companies are building something proprietary, in a world where it’s easy to build a wrapper
Ship things fast with AI tools, the world is your oyster now
“The most valuable people in the next phase of AI will be the ones who can connect creation with consequence,” Salmela says.
AI, Salmela argues, is not just a productivity layer. It’s a forcing function that exposes weaknesses in how organizations operate.
“We removed one constraint, writing code, and revealed another,” she says. “Now we have to optimize for that”
For those building careers in AI, she believes the opportunity lies in solving the problems that emerge after acceleration, not just fueling it.
“The future won’t belong to the people who ship the fastest,” Salmela says. “It will belong to the people who can keep systems running when everything moves faster than humans ever have.”
