{"id":133417,"date":"2026-07-02T00:07:39","date_gmt":"2026-07-02T00:07:39","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2026\/07\/02\/guidelines-for-respectful-use-of-ai-oreilly\/"},"modified":"2026-07-02T00:08:47","modified_gmt":"2026-07-02T00:08:47","slug":"guidelines-for-respectful-use-of-ai-oreilly","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2026\/07\/02\/guidelines-for-respectful-use-of-ai-oreilly\/","title":{"rendered":"Guidelines for Respectful Use of AI \u2013 O\u2019Reilly"},"content":{"rendered":"

<\/a>\n
<\/p>\n

\n
\n

The next article initially appeared on<\/em> Medium<\/a> and is being republished right here with the writer\u2019s permission.<\/em><\/p>\n<\/blockquote>\n

As corporations undertake AI instruments, lots of time is spent on desirous about AI insurance policies from a safety, compliance, and even cost-focused angle. However many leaders are neglecting to deal with how their groups ought to work with AI within the context of the group as an entire. This creates lots of unresolved stress, and it\u2019s time for leaders to step up and set some pointers not only for use AI in an \u201cauthorized\u201d sense however use it respectfully.<\/p>\n

Once I say respectfully, I’m not speaking in regards to the baseline acceptable office conduct (bullying, abuse, harassment, and so on.). As a substitute, I\u2019m involved that many people haven\u2019t thought of that the methods AI could make a person extra productive (actually enabling them to provide extra outputs) can have an general adverse affect on the group\u2019s productiveness. Leaders can\u2019t simply sit round and count on that their groups will know that they’ll\u2019t simply produce slop and ship it to others; when you haven\u2019t arrange an intensive policy<\/a> but, listed below are some strategies on what to cowl.<\/p>\n

Components of respectful AI use<\/h2>\n

Don\u2019t ask somebody to learn\/assessment what you haven\u2019t learn or reviewed your self.<\/p>\n

This is likely one of the commonest frustrations I hear amongst individuals engaged on AI-heavy groups. Whether or not it\u2019s code that the proprietor didn\u2019t actually hassle to grasp earlier than submitting for assessment or paperwork that they generated and didn\u2019t hassle to learn, too typically individuals attempt to steal productiveness from their colleagues by streamlining their manufacturing of labor whereas asking their colleagues to do all the high quality management themselves. It\u2019s nice to have a loop of AI code technology \u2192 AI code assessment \u2192 AI fixes \u2192 ultimate human assessment, but when the particular person prompting the AI doesn\u2019t hassle to assessment that code first, they\u2019re placing an enormous validation tax onto their teammate, who has to belief each that you just prompted effectively AND that the AI understood the context and downside effectively sufficient to get a sustainable resolution.<\/p>\n

Paperwork are a good larger temptation than code, as a result of AI is so verbose and most of us hate writing and modifying. It\u2019s straightforward to get right into a loop the place you ask the AI some questions, skim the solutions, output a doc and ship it to others. I\u2019m responsible of this myself! However what is sensible if you\u2019re skimming one reply at a time could not make for an excellent general doc, and there’s a huge distinction between answering particular person questions and writing for a human reader. Particularly, the context that you’ve got in your individual head as you’re speaking to the AI could not come out in any respect within the doc; when you don\u2019t hassle to learn it totally earlier than sending it out, you received\u2019t catch the hole in framing.<\/p>\n

Even worse, generally individuals don\u2019t even perceive what the doc they prompted is attempting to say. Are you able to describe this doc, and have a dialog in regards to the ideas it presents with others and why it is sensible? If not, you don’t have any enterprise sending it alongside with out at minimal the massive caveat \u201cThat is AI-generated and I nonetheless don\u2019t actually perceive this area, please assist me.\u201d<\/p>\n

Many individuals have reached the purpose the place they received\u2019t learn one thing an individual didn\u2019t hassle to jot down themselves, and who can blame them when so many don\u2019t even hassle to learn their output earlier than sending it on?<\/p>\n

Shorter is best.<\/h3>\n

A part of the annoyance of reviewing AI-generated work is that the AI might be painfully long-winded. AI code typically appears like tutorial code, with way more verbosity than human builders would hassle with. Add within the temptation to one-shot huge adjustments moderately than desirous about break the code down into items, and you may find yourself with stacks of thousand line pull requests. The paperwork AI produces are so thorough that one thing that ought to be 3 pages turns into 10 or 20. And for many who have absolutely embraced AI for all of their text-based interactions, you begin to see the LLM-generated wall of textual content chat messages or emails.<\/p>\n

That is, frankly, simply impolite. It goes hand in hand with not bothering to assessment your individual work, however even when for some purpose you persuade your self that you just actually did learn and edit that big PR\/doc\/message, you\u2019re nonetheless asking a lot extra of the viewers than you most likely put into the train within the first place. In relation to code, I encourage you to truthfully ask your self: If this broke at 3:00am and not one of the AI instruments had been working, would you have the ability to have a look at the PR context and the change and debug it? If not, it’s most likely an excessive amount of. In relation to a giant doc, at a minimal, have you ever at the least summarized the details up-front? If somebody is simply going to ask an AI to summarize the doc themselves, it’s best to most likely do extra work to offer that worth earlier than handing it off.<\/p>\n

Lastly, when you\u2019re writing long-winded emails or chat messages with AI-assistance as a way to painstakingly attempt to clarify one thing, maybe you truly have to have a gathering or name as an alternative. More and more lengthy textual content exchanges have all the time been an indication that folks have to cease and speak face-to-face, and AI logorrhea hasn\u2019t modified that.<\/p>\n

AI isn’t an excuse to show off your mind, or your coronary heart.<\/h3>\n

Indicators we\u2019ve switched off our brains and our hearts embody: not reviewing the AI-generated work, not taking the time to do human modifying, not breaking the adjustments down into chunks, and avoiding actual conversations via AI-mediated textual content alternate. This steerage is about respectful use of AI as a result of when you’ve got empathy in your colleagues and respect for his or her time and abilities, you’ll present them the courtesy of giving them work that you’re pleased with, that you just stand behind, that you’ve got thought via and may clarify. The AI could have produced lots of the output, however you considered all the items that wanted to be performed, and used the additional productiveness to make one thing higher: extra dependable, less complicated, effectively examined, no matter. If you end up not considering in any respect and simply mindlessly prompting, accepting output, and transferring ahead, it\u2019s a warning signal that one thing is mistaken. Maybe take some recommendation from Vicki Boykis<\/a> on including friction to your improvement course of (or regardless of the equal is of your day-to-day work).<\/p>\n

Framing these pointers<\/h2>\n

Should you determine to do that, one ultimate tip from me: Assuming your organization has some kind of firm values, it\u2019s all the time a good suggestion to name again to those values if you create insurance policies and pointers like this. It\u2019s one factor to abstractly say that shorter is best, however when you can tie that to a worth in your firm, it’s going to resonate extra strongly. For example, if I had been at Amazon I’d contemplate tying \u201cshorter is best\u201d to the management precept Invent and Simplify<\/strong>. And since shorter is best and that is already too lengthy, I depart you right here.<\/p>\n

Take pleasure in this submit? You would possibly like my books <\/em>The Manager\u2019s Path<\/a> and <\/em>Platform Engineering: A Guide for Technical, Product, and People Leaders<\/a>.<\/p>\n<\/div>\n