On AI, Ownership, and Why Nobody Wants Your Slop: They Want You

There is a tell. Every senior engineer, every product lead, and every editor who reviews a lot of documents has learned to spot it within two paragraphs. The tell is not bad writing exactly. It is the absence of a person. The sentences are clean. The structure is logical. There is a summary at the top and next steps at the bottom. But somewhere in the middle, you realize whoever submitted this was never really in the room. \ That is what slop is. Not low effort in the old sense of misspellings, sloppy formatting, or half-baked ideas. Slop in 2026 is high-effort-looking output with no one home. And it has become the defining quality problem of the AI era. \ Please use your AI. I know, I know, it is an interesting sentence coming from me. And I hear you, dear reader: 'Michal, you sharp-witted critical thinker, after all the warnings, you want me to go all in on AI?’ Yes, yes, I do. The caveat is ‘how.’ AI tools are genuinely extraordinary, and if you are not using them, you are already behind. But there is a dark version of ‘using AI.’ \ The version where you abdicate your strategic judgment to a language model and blindly ship the raw output. This approach just scales the production of low-value work and rapidly burns through stakeholder trust. Don’t do that. \ Here is what I want to talk about: the gap between what AI can generate and what your name is worth. And what happens when you close it. The Speed Is Real. The Trap Is Real. There is a common trap professionals fall into after a few months of heavy AI adoption. While daily output volume might triple, the actual accountability drops. People start churning out technical briefs and strategy documents at unprecedented speeds, but they quickly find themselves unable to defend those same documents under scrutiny. \ This happens because when a specific line is questioned, the author inevitably wonders if they actually wrote it or if an AI did. Usually, they do not know for sure. They hit approve, they skimmed the text, but they lack true ownership over the final product. \ That distinction matters more than most people want to admit. \ When you own something, you know why every decision was made. You know what you cut and why. You know what you almost wrote before you changed your mind. You know the second-order effect of the recommendation on line 7, and you know it connects to the risk you buried in the appendix. That knowledge is not in the document. It is in you. And it is what makes you defensible, trustworthy, and worth listening to. \ AI can draft. AI cannot own. The moment you confuse the two, you start shipping documents that look like yours but are not. The speed AI gives you is a gift. The trap is using that gift to produce more of something you don't actually stand behind. The Reviewer Test (They Know Where You Live) Here is a practical heuristic I have started using, and I have heard versions of it from others independently. If your document took less time to produce than it takes a thoughtful person to read and understand it, you probably shouldn't send it. \ That is not a rule about hours spent. It is a rule about the depth of thought. If reading your PRD takes thirty minutes and writing it took three, the ratio is wrong. This is not because three hours is insufficient labor, but because thirty minutes of reading implies thirty minutes of ideas, context, caveats, and judgment calls that need to be in there. And those take time to put in. Not because typing is slow, but because thinking is. \ The people reviewing your work, like your tech lead, your product counterpart, and your manager, have learned to feel this. Not consciously, always. But they know when a document is load-bearing versus decorative. They can tell when the author would fold under a single probing question. And when that happens, it doesn't just hurt the document. It hurts you, your credibility, your career, and the trust people extend to you in the next high-stakes conversation. \ There is a phrase I have started using internally: reviewers hate slop, and they know where you live. Meaning that anonymity doesn't protect you here. Your name is on the document. When someone senior spends thirty minutes reading something you clearly spent two minutes reviewing, they notice. \ The standard your reviewers hold you to has not dropped because AI exists. If anything, it has risen. Everyone now assumes you had help. The question is what you did with it. A New Workflow That Actually Works Here is what I have landed on after enough iterations to feel reasonably confident in it. Draft fast. Think slow. Use AI to get something on paper quickly. Do not make a finished document, but rather a structure, a first pass, a working prototype of the idea. Don't edit the grammar. Don't polish the sentences. Just get the shape of the thing. Act like a hostile reviewer. Stop and read it. Not looking for typos, but looking for places where you wouldn't stand behind a claim if challenged. Look for logic that sounds plausible but that you couldn't actually defend with data. Look for recommendations that AI generated because they are statistically common in documents like this one, not because they are right for your situation. Those places are where your work begins. Comment on everything you are changing. This sounds bureaucratic, but it is actually the fastest way to stay sane and keep your team aligned, especially when you are modifying someone else's work. When you make a judgment call to change a recommendation, cut a section, or add a caveat, write a comment explaining what you changed and why. Or ask AI to do so. Do not just do this for yourself in three weeks. Do it for the original author, the developers, or the stakeholders relying on that asset. Whether you are submitting a pull request for a shared codebase or updating public documentation, leaving a clear trail of your reasoning keeps the collaborative process smooth and respectful. It forces you to actually know why you made the change. It also makes reviewing radically faster and more honest. When your co-author sees a clear comment stating "changed priority here because the Q3 roadmap dependency makes this a blocker, not a nice-to-have", they can evaluate your strategic judgment rather than guessing your intent. When they lack that context, they are just evaluating whether your sentences parse. And remember, reviewers hate slop, and they know where you live. Use bullets for speed, paragraphs for ownership . There is a rhythm that works well. Capture your raw thinking in bullets as rough as you want, typos and all. Let AI clean the language up. Then go back through the full text and make sure every sentence is saying your thing in a way you endorse. The AI handles expression. You handle meaning. This keeps you honest about what you actually think, and it keeps the document sounding like a person. Specifically, like you. What to Do With the Week You Got Back Let's say you used to spend a full week producing a complete feature epic. That means drafting multiple technical design documents, writing out endless user stories, defining exact specifications, and maybe even reviewing hundreds of lines of code. Now, thanks to AI, you can generate that entire baseline in a single day. So, what do you do with the remaining four days? \ Most people simply hit "generate" again and pump out more volume. For a while, this looks like productivity. \ But the people who are actually getting better at their craft use that newfound time to transform generated slop into exceptional quality. They do not just accept the first draft. They use those four days to aggressively interrogate the output. They ask what the AI assumed, what edge cases it skipped, and how it impacts the actual user experience. They fix the underlying logic, refine the architecture, and ensure every requirement is bulletproof. \ They also use that time to have the critical conversations they previously had to skip because the deadline was too tight. They schedule that extra sync with the engineering team to map out technical constraints. They run that customer discovery call that used to feel like a luxury. They do the strategic work that makes future work better. \ You are spending the rest of your week acting as the ultimate editor and quality gatekeeper. The final result is that the uncompromising quality you expect is absolutely there, but you arrived at it much faster and with significantly less friction. \ Here is what that produces over time. An engineer or product manager whose AI-assisted output is worth substantially more than a peer's AI-assisted output. Not because the tools are different, since everyone has the exact same tools. But because one person used the speed to go deeper into quality assurance, and the other used it to go wider into mediocrity. The volume advantage of AI is temporary. The ownership advantage is durable. On the Strange Responsibility of Having Good Tools There is something almost philosophical about this moment. For most of history, the constraint on professional output was labor. Writing was slow, research was slow, and synthesis was slow. The effort you put in had a roughly proportional relationship to the quality you got out, and volume was naturally self-limiting. \ AI breaks that relationship. You can now produce the form of high-quality output very quickly. The paragraphs are there, the structure is there, and it reads as if someone who knows what they are doing wrote it. But the substance, meaning the actual knowledge, judgment, and accountability, still takes time to develop and apply. It cannot be generated. It can only be you. \ This is new. And it creates a responsibility that didn't really exist before. It is the responsibility to know the difference between the form and the substance of your own work, and to make sure you have both before you put your name on something. \ Using AI well is not a technical skill. It is a professional discipline. It is knowing when to accept what the model gave you and when to push back. It has a strong enough point of view that you notice when the document says something subtly wrong, even if it is grammatically correct. It is caring enough about your own reputation to not ship the first draft of something you didn't write. The Simple Thing at the End None of this is really about AI. It is about craft. \ Craft means you own what you make. You know why every decision is there. You can defend the thing in a room full of smart people who disagree with you, not because you memorized talking points, but because you thought it through. \ AI is a tool. An extraordinary one, and probably the most powerful tool most knowledge workers have ever had access to. But a hammer doesn't make a carpenter. What makes a carpenter is judgment about where to drive the nail, why this joint and not that one, and what the wood is going to do when the weather changes. \ Use AI. Use it aggressively. Use it for the drafting, the formatting, the boilerplate, the translation, the summarization, and the ten other things it is genuinely better at than you are. Let it handle expression so you can focus on meaning. \ But know your output. Read it. Challenge it. Change it. Put yourself in it. Add your context, your opinion, and your knowledge of this specific situation that no model trained on the whole internet could possibly have. \ Because here is the thing that is still true, and will keep being true: nobody wants your slop. We want you. And right now, you have more time than ever to show up. \