I believe in AI. But my AI assistant gave me more work than it could finish

I awoke to a daunting to-do list - 156 items in all. As a writer and entrepreneur, such a number felt like the productivity system had spun out of control. Tasks were multiplied, overlapping and overlapping again, each item carrying the weight of uncompleted work. I had sought AI assistance to streamline my workload, expecting it to ease my burden, not intensify it. An assistant can certainly generate plans and checklists, but it often produces more tasks than it resolves.

The key issue lay in the blurry line between progress and completion. A well-crafted plan can be incredibly motivating, yet it can also lull us into a false sense of progress without confronting the core of the task at hand. A polished draft or a filled calendar can provide a false sense of accomplishment while leaving critical steps undone.

A simple completion check-off can mask the fact that the real work remains undone - the email still waits to be sent, the application still needs submitting, the client's needs still remain unmet.

To address this, I have implemented several changes. First, I have set a realistic daily limit of 25 meaningful tasks, ensuring I maintain realistic capacity and don't get overwhelmed by an unrealistic to-do list. I also aim to separate related tasks into one clear execution item, rather than allowing them to proliferate as separate obligations. This better reflects the true scope of the work and reduces the clutter of unnecessary tasks.

Second, I have insisted on verifiable completion. If an assistant claims an email was sent, I must be able to see the proof. If it says a piece was published, I need the live link. This prevents duplicate work and helps me maintain an accurate record of what has been accomplished.

Third, I have made a point to identify specific blockers. Rather than simply stating "unable to complete," I now note the exact reason - whether it's a missing answer, a technical issue, or a lack of access. This allows me to address each obstacle directly rather than letting it accumulate and create further complications.

Finally, I have started tracking outcomes. By monitoring how many pitches were sent, which replies were needed, which deliverables were accepted and which remain unfinished, I can gain a clear sense of the actual work completed versus what still needs to be done. AI can certainly be a powerful tool in managing this process, reconciling records and handling routine steps.

However, it's essential that we guide its actions and ensure it acknowledges when more work remains unfinished. I still believe in the power of technology to make work easier and possibilities bigger. But as my to-do list, I want it to stop recruiting new tasks when the real work is already clear.