One of the most uncomfortable things about working with AI is that it does not just expose what the tool can or cannot do. It often exposes what we were never clear about ourselves. The friction we experience is not always a sign that AI is failing. Sometimes it is a sign that our instructions, decisions, and workflows were already costing us time long before AI entered the picture. That is why this matters so much for teams trying to save time. AI does not only accelerate work. It also acts like a mirror. And what it reflects back to us is often the hidden source of delay, vague thinking, unclear expectations, inconsistent handoffs, and avoidable rework that were already shaping our cycle times. ------------- The Tool Did Not Create the Confusion ------------- A common reaction to disappointing AI output is to blame the tool immediately. The answer was too generic. The draft missed the point. The summary left out something important. The recommendations felt disconnected from the real need. Sometimes that criticism is fair. But other times the real issue is more revealing. The output is weak because the input was never clear enough to produce strong work in the first place. This is not just an AI problem. It is a work design problem. Many teams operate with instructions that are functional enough for humans to patch together socially, but not clear enough to stand on their own. A manager says, “Put together something polished for leadership.” A teammate asks for “a quick update” without defining what matters. A project brief contains goals, but no decision criteria. A task gets assigned with urgency, but without enough context to reduce ambiguity. Humans often compensate for this through intuition, back-and-forth, and experience. AI cannot compensate in the same way. It reflects the ambiguity more directly. That is why AI can feel frustrating at first. It removes the illusion that the request was clear. It shows us, very plainly, how much of our normal workflow depends on people filling in blanks that were never explicitly addressed. When that happens, the tool is not introducing confusion. It is surfacing confusion that was already there.

We tend to think unfinished work is a discipline problem. We assume the issue is distraction, procrastination, or lack of follow-through. But in many cases, the deeper problem is structural. Modern work is built around interruption, fragmented attention, and constant switching, which means a surprising amount of our time is lost not in doing work, but in repeatedly returning to work that was never fully closed. That is why the half-finished task has become such an expensive unit of modern work. AI can help here, not just by producing faster outputs, but by reducing the time cost of interruption, re-entry, and stalled momentum. ------------- The Real Cost of Half-Finished Work ------------- Most teams do not struggle because nothing gets started. In fact, plenty gets started. Drafts begin. Plans are outlined. Emails are opened. Research is gathered. Decisions are discussed. Notes are captured. The real problem is that a large share of this work remains suspended in a partially completed state, waiting for the next block of attention that may or may not come soon. That suspension carries a time cost. A half-finished task is not neutral. It continues to occupy mental space, creates uncertainty about status, and increases the friction of getting back into motion. The next time we return, we are rarely able to pick up exactly where we left off. We need to remember what we were thinking, what was already done, what still matters, and why the task felt important in the first place. That means the total cycle time of a task is often much longer than the actual work required. A report may only need 45 focused minutes, but if it is started and stopped four times, the re-entry cost can stretch it across two days. An email may take five minutes to send, but if it sits half-written while other priorities intrude, it becomes one more unfinished thread draining attention in the background. This is one reason people feel productive and behind at the same time. Their days contain activity, but not enough closure. They are surrounded by motion, but starved of completion. AI becomes useful here because it can help reduce the cost of resuming, structuring, and finishing work that would otherwise remain stuck in fragments.

We often talk about time as if it only lives on the calendar. We count meetings, deadlines, deliverables, and hours worked. But some of the most expensive time loss in modern work never appears in a schedule at all. It lives in the background, in the mental effort required to keep unfinished work active in our head. That is where AI can become surprisingly powerful. Not just as a tool for output, but as a way to reduce the hidden time tax of carrying too much unresolved thinking at once. ------------- Where Time Leaks Before Work Even Starts ------------- A lot of people assume they need better time management when what they really need is less mental carrying. We do not just spend time doing work. We spend time remembering what needs to be done, revisiting half-formed ideas, holding open loops in memory, and trying not to lose important details before we have a chance to act on them. That overhead is real work, even if it does not look productive from the outside. Think about a normal day. We may have a proposal to finish, a follow-up email to send, a team decision we still need to make, a new idea for a process improvement, and three conversations that require thoughtful replies. Even when we are not actively working on those things, part of our attention stays attached to them. We keep mentally rehearsing, “Do not forget that point,” or “I need to circle back to that,” or “There was a better way to explain that.” That constant background processing drains energy long before the task itself is completed. This is one reason people end a day feeling busy but strangely unfinished. The issue is not always a lack of effort. It is that attention has been fragmented across too many mentally open loops. The brain becomes a storage system, a reminder system, and a drafting space all at once. That creates invisible cycle time. It slows the path from thought to action, from task to completion, and from idea to value. In that sense, the real time leak is not just workload. It is unexternalized workload. The more work we hold in our head, the more time we lose to friction, context switching, and re-entry. AI matters here because it can help us move thinking out of our head and into a form we can work with faster.

This week, I cover the new model releases from OpenAI including GPT-5.3 and 5.4, plus I show off the new Project Sources update for ChatGPT. I also break down the many releases from Google as they condense their AI tools down into the products that work, talks briefly about OpenClaw competitors and Anthropic's latest releases, and more. Enjoy!

Onboarding is not an HR event. It is a time-to-value challenge. Every week a new hire or newly promoted teammate spends confused is a week of delayed impact, extra interruptions, and hidden rework. The cost is not just their time. It is everyone else’s time spent answering the same questions. AI can dramatically reduce time-to-competence when we use it as a knowledge multiplier and a private tutor. But the biggest benefit comes when we pair AI with clear artifacts that capture how we work. ------------- Why Onboarding Takes So Long ------------- Onboarding drags because knowledge is scattered. The process lives in someone’s memory, in old docs, in Slack threads, and in unspoken norms. New people do not just need information. They need context, priorities, and examples of what “good” looks like. When that is missing, they ask more questions, and they make avoidable mistakes. Those mistakes create rework, which slows them down and creates frustration. Meanwhile, the team gets interrupted, which increases context switching and slows everyone down. Time outcome: the onboarding problem is really a handoff latency problem at scale. ------------- Insight 1: The Fastest Onboarding Comes From Patterns, Not Pages ------------- Many teams respond to onboarding gaps by writing huge documentation. That often fails because it is hard to maintain and hard to consume. What actually helps are patterns: templates, checklists, examples, and definitions of done. New hires do not need everything. They need the 20 percent that lets them deliver the first 80 percent of value. AI helps us extract patterns quickly. We can feed it examples of past work and ask it to identify structure, tone, and success criteria. Then we turn those patterns into repeatable assets. Time outcome: faster time-to-first-independent-deliverable. ------------- Insight 2: AI as a Private Tutor Shrinks the Learning Curve ------------- New hires hesitate to ask questions because they do not want to look unprepared. That hesitation costs time.