This is the productivity paradox of our moment. AI doesn't necessarily "save time" — it raises the bar for what "enough" looks like. The work expands to fill the new capacity. And because it feels faster, nobody notices until burnout hits.
This absolutely resonates with both my personal experience and what I'm hearing. Not necessarily the piece about giving up on tools that slow them down - everyone I've spoken with are willing to try. In my mind the issues people are faced with are:
a) How do I talk to this agent thingy to get what I want out in a way which I'd be happy with?
It's a whole new paradigm/tooling/skill which everyone is a junior with and makes seasoned professionals either feel excited at learning something new or like they're falling behind seeing others get masses of things delivered. In a time-bound situation it's more likely to cause frustration getting over that learning curve and people fall back into old habits easily when under strain. Add into that the failure modes for AI are mostly hidden away from where the interaction is (you chat in text, the failure is in the code somewhere) and it adds an additional layer of complexity that people are learning how to deal with. Doing the training and practice will help here, but people need to accept there will be times when the AI just doesn't work the way you want.
b) There's now so much more to review!
Agents are capable of outputting a greater number of larger PRs. Engineers are still responsible for the code that's delivered so they need to review what the agents are producing. In addition to that, most teams require 1 or 2 peer reviews so not only the original engineer needs to spend time on each PR, but so do 1 or 2 other people. The scope and context of what needs reviewing is much greater that before with a significantly (orders of magnitude) faster cycle. As we make more use of AI code review tools to do initial passes and we learn how to control the tools to give us more succinct PRs this will get better.
c) I'm still responsible but didn't build this
Most of our people are not used to delegating the execution of a task away while keeping responsibility for the output. It's like everyone becomes a team leader for a squad of super-knowledgable junior/mid-level coders (NOT engineers) overnight. It's a mind shift that will take some people longer to get past than others.
It's as if everyone became juniors at the tooling overnight with the same, or increased, expectations around delivery velocity. It adds an additional layer of communication which is almost always a friction point (even minor) when the only form of communication is text.
We'll get past it as people get more familiar with the tooling and learn how to work with them sustainably, but it's definitely something to be considerate of as we move through Q2 and into Q3.