Sorted but stuck

The gap between knowing what to hand to AI and actually doing it

I ran a workshop on agility and resilience last week. The room was sharp and curious. Almost everyone had taken an AI course. They wanted this to work.

Then someone said the quiet part out loud. They had a list of things that could maybe shift to AI. They just couldn’t move past the list. Picking what to try felt fuzzy, especially when the work was shared with other people. And sometimes AI added cycles instead of removing them, so they churned on a task longer than if they’d just done it themselves.

The quiet truth I keep observing is a friction point between knowing and doing. Not a lack of interest. Not a lack of training. People know roughly what could move. They get stuck on how. I’ve started calling it being sorted but stuck.

Why the list isn’t enough

A list of candidate tasks is a map; it captures the potential. Now, the impactful part is turning that insight into a clear, actionable step that enhances your workflow and makes your week better.

The rapid pace of change is creating an even bigger sense of urgency. Prosci found that 22% of employees struggle with AI’s learning curve, and that the half-life of AI skills is roughly three to four months. So the course you finished in the spring is already going stale. That’s not a reason to quit. It’s a reason to stop treating this like one big leap and start treating it like small reps you repeat.

Here’s the part that should make all of us pause. On the Big Technology Podcast in late May, the hosts shared a sobering number. “Of people trying to actually use these tokens to accomplish things, 82% of that use is not translating into shipped products that reach real users.” Eighty-two percent. Effort going in, yet the measurable, finished output is minimal.

I don’t think that’s a people problem. I think it’s a guidance problem. 

Two tools, two jobs

So I built two small things that work as a pair. One handles the what. The other handles the how.

The first is The Work Sort Canvas. You brain-dump a normal week, then drop each task into one of three buckets. “My work” is human only, where your judgment and relationships live and AI stays out. “With me work” is you plus AI, for first drafts and second opinions, with you in the driver’s seat. “For me work” is routine and rule-based, the stuff you hand off and check. The canvas gives you the what.

The second is the Work Sort AI prompt. It addresses the how. Paste it into Claude (or your GPT of choice), attach your filled-in canvas, and it interviews you before it does anything. It proposes an approach and waits for your go-ahead. You stay steering the whole time. 

Measure the right thing

When you try it, please don’t count prompts. Prompt counts are a vanity number. Watch two things instead. How the work felt: less overwhelm, more control. And how the work turned out: sharper thinking, fewer slips, more room for the human parts. That’s the signal that you’ve moved something real, not just generated more tokens.

Stepping back

We’re at a tricky moment. A lot of people are adaptive and willing to change. What’s thin is the guidance. And without it, all that effort drains into that 82% that never lands. My hope is that these two tools shift even a little of that into something genuinely useful, one task at a time, without piling more change on top of you than you can carry.

My question for you: where are you sorted but stuck right now?