Owl Things Considered: Issue 8

Common writing flaws that confuse AI summaries (and humans), why we admire design studio Love at First Try, and this issue's dopamine dose

Published

August 17, 2026

Knowledge management tips: Use your words (or the robots will get confused)

Lisa Simpson in a tie-dyed shirt, cap, and granny glasses looking at her reflection in a discarded Jiffy Pop container

Common writing flaws that mess up AI summaries (and human comprehension)

Docs genius and human delight Casey (CT) Smith ran some experiments recently while working on her upcoming book, testing which common writing flaws cause AI the most trouble when summarizing or answering questions about your docs. She shared the code she used to run the tests, along with much more detail of the flaws she tested and her findings, if you want to dig into the methodology (which I super recommend you do!).

Here’s what stands out to us from her findings:

Quantifier vagueness and scope hedging caused the most issues, and weren't "fixable" in any meaningful way. Replace vague quantifiers like "periodically" or "an error" with specifics, like "every 90 days" or "a 404 Not Found error," and cut vague scope qualifiers like "generally," "where applicable," "in most cases," or "depending on your configuration." Left in place, the AI can invent a specific value that was never in your original document or will mirror the vagueness.
Undefined jargon caused the most visible failures. Always define abbreviations, acronyms*, or internal terminology on the page. Otherwise, the AI parrots the term back without explanation, hallucinates a plausible-sounding but incorrect definition, or refuses to answer altogether. (A glossary feature is one way to quickly mitigate this issue!)
Ambiguous pronouns confuse the AI when a document covers multiple features or concepts. Structure paragraphs and sections so it's clear what "it," "this," or "they" refers to, or the AI may confidently answer about the wrong "it" or give a vague, unactionable answer.
Passive voice was technically fine, but it led the AI to generate passive-voice-structured answers in return, which may be something you want to avoid. See Marcia Reifer Johnston’s wonderful blog post about how to avoid passive voice.

These are good general writing habits regardless, but they matter even more if your docs are feeding an AI chatbot or any other AI-powered tool.

* A few months ago, when I was using Claude to help me plan out a format for our Product Newsletter I mentioned TNBTW (The Not-Boring Tech Writer podcast), using just the acronym, and it delightfully referrered to it as the There’s Nothing Better Than Waffles podcast. No biggie when you’re just using it for planning, but yes biggie when you’re working on product docs!

Businesses we admire: Love at First Try

SaaS companies, and the companies who serve them, don't always have the best reputations. And that's not improving with the tidal wave of AI currently washing over all of us. It's in this environment that the most human of companies stand out. Careful, thoughtful design and service feel more valuable now than ever.

Love at First Try (LAFT) is an in-house product design team for B2B SaaS companies, founded by Jim Zarkadas. KnowledgeOwl has worked with them for a few years now, and the impact on our product has been beautiful. With a product as feature-heavy as KO, surfacing and explaining everything we offer to our customers is a genuinely huge undertaking. LAFT has helped us work towards that, again and again, without ever losing sight of the people on the other end of the screen.

They're completely on top of where software and design are heading, from AI-native workflows to code-based design systems, but they never let any of it override their judgment. New tools don't replace their taste. They use them in service of it. That distinction matters more by the day

They've also helped KnowledgeOwl find our voice. Marketing used to be a dirty word around here, something we left almost entirely to word of mouth. LAFT showed us how to present ourselves in a way that's authentic and creative, marketing that sounds like us, instead of like marketing fluff.

What I love most about LAFT is how they hold two things at once: real talent and real humility. Jim has asked me, more than once, for painting and color theory recommendations. I'm an artist myself, and there's something wonderful about a founder in the business of visual craft who stays genuinely curious about craft outside his own lane. That openness runs through the whole team. They produce work that's honestly beautiful, and they're just as willing to pivot the moment something isn't right. They operate without ego or defensiveness. They are people who care about getting it right more than being right.

They also seem to understand something a lot of vendors miss: they're not only serving us, they're serving our customers, too. Every recommendation is filtered through what will help the people using KnowledgeOwl, not just what will make KnowledgeOwl look good. That's a values alignment you can't fake, and it's a large part of why our relationship continues to grow with them.

As AI makes more and more things easy, it can't replicate the nuance of people who are genuinely curious, who listen before they act, and who are as good to work with as they are good at their work. That combination—talent paired with care—is becoming the actual differentiator. Being really good at what you do while being really good to the people around you isn't a soft skill anymore. It’s becoming the whole shebang.

Dopamine dose

Some warm fuzzies to brighten your day

gif of a white dog swimming in a sun speckled river

Here are this issue’s day-brighteners:

Song to smile to:

If I ever need to add a little extra pep to my step any song by this absolute icon will do, but Fiesta by Raffaella Carra is truly a gift for the ages.

Heart-warming animal alert:

The earthquake in Venezuela was unthinkably tragic, and nothing about this should distract from the loss so many families are living through right now. Alongside the countless humans who've stepped up to donate, volunteer, and rebuild, one particular story broke through: Tsunami, a border collie once abandoned on the streets of Caracas, has spent his career searching for survivors trapped in disaster zones from Turkey to Syria to his own hometown. In this response, he and his handler helped rescuers find 13 people alive beneath the rubble, reportedly his final mission before retirement.

Who we're giving to:

At KnowledgeOwl, we have a community commitment that we call 2% for People and Planet.

Beyond our commitment to 1% for the Planet, we also give another 1% to a charity of a different owl's choosing each month. This program allows us to help many different people in many different communities.

If you're looking for some donation inspiration, you might think about giving to organizations we've given to recently (or starting a similar program at your company!).

In May 2026, Marybeth, Chief Executive Owl, donated to WWOZ, a listener-supported, non-commercial radio station for New Orleans and the surrounding community. They have to make up a $400k shortfall each year now that the government dissolved the Corporation for Public Broadcasting.

In June 2026, Pete, Chief Technical Owl, donated to the Mississippi Wildlife Rehab, an org dedicated to rescuing, rehabilitating, and releasing native wildlife.