You have 1 article left to read this month before you need to register a free LeadDev.com account.
Key takeaways:
- AI moves the bottleneck, it doesn’t remove it.
- Generation got cheaper, but judgment didn’t. The bottleneck shifted to the layer where staff+ engineers operate.
- More capability means more demand: AI expands the surface area where you can contribute.
There is a strange thing that happens when AI starts working just right. At first, it feels like magic. You feel like you have a new super power. The blank page is less intimidating and the first draft of the design doc is easier to get to. The unfamiliar codebase is easier to explore. The scattered thoughts in your head become something you can organize and share with a team.
All of that is real. I have felt those gains directly. As a staff+ engineer, I now use AI constantly. I use it to explore systems I do not know well, summarize tradeoffs, prepare technical proposals, compare documentation against code, reason through architecture options, and turn messy thinking into something other people can react to. There are days when I have several agent sessions open at a time, each attached to a different project or workstream I’m working on.
If AI is making you more productive then, you would expect to have more time for deep work and a calmer calendar. However, I feel busier than ever. If AI is making me more productive, shouldn’t I have more breathing room? Well, not necessarily. AI does not simply remove bottlenecks. Often, it moves them.
I think many staff+ engineers are starting to run into this paradox. AI reduces friction in the work, but it also expands the surface area where we can contribute. It produces more drafts, more reviews, and more decisions.
For many organizations, the bottleneck is moving toward the exact layer where staff+ engineers operate: verification, architecture, alignment, and judgment.
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
The work didn’t disappear, it just changed shape
AI lowers the activation energy for work. It gives me a starting point. It helps me move from “I have a pile of thoughts” to “here is something a team can inspect, challenge, and improve.” That is productivity, but it is not the whole story.
In practice, the work changes shape. This is especially true for staff+ engineers because much of our work is elastic. There is always another architecture review that could use input, another team that needs help clarifying direction, or another cross-team dependency that nobody fully owns.
AI makes it easier to contribute to more of those conversations.That is extremely valuable, but there is a tradeoff. When you become faster at entering workstreams, the organization may simply route more workstreams through you.
The bottleneck moves to judgment
Before AI, there were many things I might not have been deeply involved in because the activation cost was too high. AI changes the calculus. It allows me to get enough context to participate in more places. It lets me maintain parallel workstreams with less friction, and, especially, it makes it easier to produce artifacts that move conversations forward.
For a lot of engineering work, AI reduces the cost of generating output – code, documentation, tests, migration plans, diagrams, summaries, or options. However, the bottleneck was not always generation.
Often, the bottleneck was knowing what should be generated in the first place. What problem are we actually solving? What constraints matter? What tradeoffs are acceptable? How does this fit with the broader system? Those are judgment questions, and as generation gets cheaper, judgment gets more valuable.
This is one reason staff+ engineers may feel busier in the AI era. The organization can now produce more possibilities – more code, more proposals, and more technical paths that are plausible enough to deserve consideration.
However, plausible is not the same as good. A weak AI answer is easy to reject. A plausible AI answer is harder to reject. It may follow local conventions. It may look like something an engineer would write, but it can still hide the wrong assumption, miss an architectural constraint, create a security issue, or solve the problem in a way that quietly raises the cost of every future change.
That means someone has to evaluate it. Someone has to ask whether the approach fits the system. Someone has to understand the consequences. In many organizations, that someone is often a senior, staff, or principal engineer.
AI is very good at producing artifacts that look like progress. Staff+ engineers have to help organizations distinguish between artifacts and outcomes. A design doc is not valuable because it exists. It is valuable if it improves the decision. The artifact is not the value. The better decision is the value.
More like this
The staff+ role becomes more important, but also more demanding
The value of staff+ engineers was never just “this is the person who writes the hardest code.” Staff+ engineers create leverage through direction, alignment, architecture, mentorship, review, cross-team influence, and the ability to reason about systems over time. AI turns up the volume on those responsibilities.
If code is easier to produce, then deciding what code should exist matters more. If teams can move faster, then making sure they are moving in a coherent direction also matters more.
AI increases the number of situations where our judgment can be useful. It gives us more ways to help, which means we need more discipline in deciding where to help.
A staff+ engineer with AI can become a force multiplier. A staff+ engineer without boundaries can become a very efficient bottleneck.
AI raises the ceiling. Organizations have to raise the floor
I remain optimistic about AI in engineering. It has changed how I work. It has made me faster in real ways. It has helped me stay involved in more conversations, understand systems more quickly, and produce better starting points for collaboration.
Yet, the fact that engineering can do more does not automatically mean the organization is healthier. We need to build an engineering operating model that can handle AI-assisted speed. That means better specs, better review practices, better architecture boundaries, better decision records, and better ways to define done.
It also means being more intentional about where staff+ attention goes. Not every AI-generated artifact deserves a staff+ review. Not every plausible idea deserves organizational energy. When the cost of starting work drops, the discipline of choosing work becomes more important.
AI raises the ceiling on what an individual engineer can do, but to benefit from that, organizations also need to raise the floor: better engineering discipline, better quality systems, better verification, and better judgment about what work is worth doing.
Otherwise, AI can make us more productive and more overwhelmed at the same time. AI can help us generate more. Our job is to make sure we are generating the right things, verifying them well, and not confusing motion with progress.
The goal is not to be busier. The goal is to make better, safer changes faster, without quietly making the future harder for everyone who has to live with the system after the decisions are made and the PR is merged.
