The other evening, at the pre-dinner drinks before the Casaleggio Associati event, I was chatting with the good Maurizio Benzi about the new personal work methodology that’s emerging from AI tooling.
What’s changing isn’t really how much we produce — it’s how we produce it.
Until recently, we interacted with productivity tools synchronously: ask, wait, receive, evaluate, repeat. Increasingly, the pattern looks like this instead: assign, move on, come back later.
It sounds like a subtle distinction. In practice, it changes everything. When you delegate to a process — agentic or otherwise — and retrieve the result afterward, your job is no longer simply to produce output. It’s to orchestrate output-production pipelines and make sense of what comes back.
Agents work in the background while you move on to something else. That eliminates dead waiting time, which you can now spend spinning up more work.
Look at it systemically
The consequence is that you’re no longer managing sequential tasks — you’re managing parallel pipelines. Every so often you loop back to each one, polling the machine: where are we? and which results are ready for review?
In this sense, productivity is no longer tied to time spent. It’s tied to the number of decision cycles you can close. Whether you’re building documents, handling daily tasks, or writing code, you no longer start from a blank page — you start from intermediate outputs you didn’t primarily produce. You’re the reviewer and the final decision-maker, not the author.
Which means the bottleneck is still us — the managers who have to make the calls. With the added wrinkle that AI, by efficiently parallelising both our work and our teams’, floods the mill with far more water than it was built to handle. The risk isn’t falling behind on production. It’s decision paralysis, the kind that comes from context-switching at this kind of velocity.
The mental inversion
This forces a mental inversion: where we once optimised for execution time, we now optimise for review time. Where value used to live in the volume of production, it now lives in the clarity of intent and the quality of output selection.
Working this way demands new habits.
First, you need to know how to define clean, delegatable tasks — for both wetware (humans, that is) and software. Second, you need to manage multiple workstreams in parallel without losing the thread on any of them. Third, you need to build deliberate “results collection” moments and the critical analysis that goes with them.
From pipeline to portfolio
In practice: you stop working in a line and start working in a portfolio.
People who keep using agents as synchronous chat interfaces will capture a fraction of the available advantage. Those who treat them as an asynchronous workforce will start to genuinely scale.
At that point, the constraint won’t be how much we produce.
It’ll be how fast we can decide.