Over the last few months, most of my time has been spent inside financial firms helping people adopt AI, and many of the workers I'm seeing aren’t trying to redesign their jobs. They’re trying to finish their work, faster.
Those of us building with AI tend to assume other workers will approach things in the same way we do, once they have the tools. We look at our own work as something to take apart and rebuild, and when the people around us do that too, it's easy to mistake what we do for a preview of how the rest of the workforce will behave.
Take someone who has worked out a way to automate a reporting process. As a builder, I expect them to then start questioning why the report exists in the first place, or how to rebuild the process around an agent. It took me a while to understand that, for many of the people I work with, the faster report was already the outcome they wanted.
The idea that AI will turn everyone into a builder assumes that technical difficulty was holding people back, so making software easier to build will release all those waiting ideas. Coding was a real constraint for some people, but building takes an inclination to question how work gets done, as well as the ability to change it. Access to the tools doesn't automatically create either.
My prediction is that firms will get a lot more done without changing much of how they work, and that organizations will change much more slowly than technology. Suggestions that firms must immediately rebuild around agents or be replaced underestimate how much has to happen for a useful tool to change a business. A worker doesn't need to build a tool to have their job changed by it, but introducing that tool across a firm can mean procurement, security review and agreement between departments, and a model release doesn't resolve any of that.
Some existing workflows will even last longer because of AI. Consider a recurring report assembled from systems that don't talk to each other. If a model helps draft the commentary or checks the figures, the report will take less effort to produce, while the meeting it feeds and the systems behind it stay the same.
For the person producing it, that can be a perfectly good result, but for someone proposing a replacement, it can make the case harder: The old process is now less painful, while replacing it will disrupt other people's work. An improvement that builders see as the beginning of a redesign can become a reason to leave the process alone.
Most workers will therefore get real value from AI while carrying on with familiar work. The faster report is a success; expecting them to redesign their jobs says more about how builders think than about what those workers want.
