A team can start using an AI tool without changing much about its work. People keep the same responsibilities and use the tool to finish some tasks faster.
Over time, those responsibilities may change. The team may give larger tasks to the tool or take on projects it could not maintain without it. At that point, losing access could leave work that nobody is ready to take over. The date the tool was released tells us little about when that dependence began.
Electric motors changed how factories could be arranged
Early factory electrification gives a concrete example of work changing around a tool. A factory could replace its steam engine with one large electric motor while keeping the same system for driving its machines. The motor turned a long shaft, and belts carried power from that shaft to individual machines.
The machines still had to be placed where the belts could reach them. Replacing the engine alone did not remove that constraint.
Giving each machine its own motor allowed a different layout. Machines could be placed in the order needed for production, so materials could move more directly from one step to the next. A Federal Reserve Bank of Chicago history uses this transition to explain why the larger productivity gains from an invention can take decades to appear.
The installation date records when the factory acquired electric power. Understanding the later gains also requires knowing how the factory used that power to change its work.
Follow the changes in people's responsibilities
With AI, the changes around the tool are worth examining too. A coding agent might write an implementation while people continue to plan, review, and maintain the software as before. A team could also give the agent more responsibility, change how it reviews code, or commit to maintaining more projects.
Those are changes in how the work is organized. They may help the team, or they may leave it with responsibilities it cannot reliably meet. The fact that a task is now delegated does not tell us whether the new arrangement is better.
Some effects of a technology can emerge much later. In a conversation about ancient DNA, David Reich describes natural selection becoming more intense thousands of years after farming began, as human environments continued to change. The underlying study examines changes in genetic variants across Europe and western Asia. This is an example of delayed consequences, not evidence about AI productivity. The factory history is the closer comparison for a team deciding how to organize its work.
Check which work depends on the tool
Imagine the team loses access to its AI tools for a week. Start with the work already planned. For each task, identify someone who could complete it, check the result, and deal with any failures.
Taking longer is an expected consequence of losing a useful tool. A more serious gap is a task that nobody can take over, or a project that the team no longer knows how to maintain. That gap points to a responsibility that needs an owner and a workable backup plan.
The reason for the gap matters. Poor documentation or lost familiarity can make a task difficult even if the team has not changed how it assigns work. Compare responsibilities and project commitments before and after the tool arrived. This helps show whether AI created a new dependence or made an existing weakness easier to overlook.
A team may decide that the new division of work is worth keeping. It should still be able to say which tasks depend on the tool, what the tool is allowed to change, and who checks the result. Those details provide a clearer basis for a decision than the speed of the next generated draft or patch.