Top Microsoft Advisor: "Coding Is Cheap, Software Is Expensive." You're Focused on the Wrong Thing
TL;DR
Suzanne Daniels says the productivity story is too small — early AI adoption centered on code completion and “55 times faster” marketing, but she argues the real gains now are better quality, better outcomes, and a faster pace of innovation, especially since developers only spend a small slice of the day actually coding.
“Coding is cheap, software is expensive” is the core mental model — Daniels’ point is that the value was never the typing itself; it was understanding the customer, defining the right solution, and turning business ambition into durable software.
Junior engineers may be undervalued right when AI makes them more important — Daniels says research shows juniors adopt AI more readily, are less boxed in by old definitions of developer work, and can become the change agents if seniors codify best practices and guardrails.
The winning teams won’t be lone geniuses with copilots — they’ll be organizations with culture and constraints — Daniels keeps coming back to platform engineering, shared standards, guardrails, and enabling others, arguing that AI performance means making the whole team successful, not just wearing the superhero cape yourself.
The advice for engineers is to widen, not narrow, their skill set — she recommends learning product, UX, security, systems thinking, specifications, and classic principles like SOLID because in an agent-heavy world, defining constraints and making intentional software matters more than raw code output.
Tool choice matters less than enablement and experimentation — Daniels says most AI coding tools overlap on 80% of functionality, but companies fail when they just hand out licenses; developers need training, internal best-practice sharing, and permission to experiment daily or they risk falling behind.
The Breakdown
Beyond the “55x faster” hype
Suzanne Daniels opens by pushing back on the dominant AI narrative: productivity. She says the industry got hooked on flashy claims like coding “55 times faster,” but that misses the point because most developers only spend a fraction of their day actually writing code. For her, productivity is just the first rung — the bigger story is better outcomes, higher quality, and accelerating innovation.
The real job was never just typing code
Daniels gets personal here: what she loved about software engineering was understanding the problem, knowing the customer, balancing business ambition, and shaping a solution. Her punchiest line lands hard: “The value was not in the coding… the value was in finding that solution.” That’s where her bigger thesis comes from — coding may get cheaper, but software, in all its complexity and longevity, is still expensive.
Why juniors may be the surprise winners of the AI era
The conversation turns to the job market anxiety around junior engineers, with examples of grads sending thousands of applications and getting nowhere. Daniels says that logic is short-sighted: the economics of hiring only seniors doesn’t work, and research suggests juniors are actually more likely to adopt AI, handle ambiguity, and stretch beyond rigid job descriptions. Her take is almost a reversal of the current fear cycle — juniors could become the change agents, while seniors provide the tribal knowledge and guardrails.
Culture, not raw horsepower, determines whether AI teams work
One of the most memorable stretches is Daniels imagining a small team — one senior, two juniors, and a bunch of agents — which some people would call a nightmare scenario. Her answer is basically: only if the organization has no guardrails. If senior engineers codify best practices, platform teams reduce friction, and the culture rewards collaboration and experimentation, then neither enthusiastic juniors nor agents have to “run wild.”
What engineers should study now that code is abundant
When asked what people should focus on, Daniels doesn’t say “learn the hottest framework.” She points back to fundamentals: specification work, systems thinking, SOLID, product sense, UX, security, and all the old principles that suddenly matter again because someone has to define the constraints correctly. Her framing is clean and memorable: don’t just produce code — produce software.
Will programming languages matter less — and apps themselves change?
The interview then zooms out into a more speculative, fun zone. Daniels wonders whether in a few years natural language might effectively become the interface for building applications, and whether users will even interact with classic apps the same way when agents can buy insurance or groceries on their behalf. The JavaScript joke lands, but her bigger point is that the entire concept of software, not just languages, may be up for reinvention.
Generic AI slop, open source pressure, and the need to add real value
Daniels agrees with the idea that if everyone uses AI to generate the same baseline output, your value has to come from what you add on top. She points to LinkedIn as the obvious example of samey AI writing, then connects that to software and even to open source monetization: if someone can replicate 90% of a premium feature from docs and a demo, companies need to offer something truly differentiated. Still, she’s hopeful open source remains strong — and says the industry needs to do a better job explaining why it matters and why people should contribute back.
Tooling is becoming commoditized, but enablement is the real moat
On developer tooling, Daniels is blunt: most tools are more similar than different, and if they’re not today, they probably will be in a few months. The real failure mode is when companies hand developers a license and call it transformation; people need training on prompts, constraints, agents, standards like MCP, model selection, and secure usage. Her final practical advice is sharp: if your company won’t let you experiment with these tools day to day, that may be a legitimate reason to leave — because this craft is changing too fast to sit still.