<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Kristina Rivers on Medium]]></title>
        <description><![CDATA[Stories by Kristina Rivers on Medium]]></description>
        <link>https://medium.com/@kristinastoney?source=rss-e668e7725b62------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*289gajaeqmi2H0qm.jpeg</url>
            <title>Stories by Kristina Rivers on Medium</title>
            <link>https://medium.com/@kristinastoney?source=rss-e668e7725b62------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 22:40:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kristinastoney/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Vanilla Effect]]></title>
            <link>https://medium.com/@kristinastoney/the-vanilla-effect-cd338c982887?source=rss-e668e7725b62------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd338c982887</guid>
            <category><![CDATA[communication]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-companion]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Kristina Rivers]]></dc:creator>
            <pubDate>Fri, 06 Mar 2026 00:54:03 GMT</pubDate>
            <atom:updated>2026-03-06T00:54:03.017Z</atom:updated>
            <content:encoded><![CDATA[<h3>What is your favourite ice cream flavour? Mine is mango. Ask five different people and you will get five different answers, because they are five different people with unique taste, history, and flavour.</h3><p>Now here is what happens when you ask AI to improve your content.</p><p>You write something. A strategy document, a piece of communication to your team… a LinkedIn post. You use AI to help you refine it. You type: <em>I am Kristina. I like mango ice cream.</em></p><p>The algorithm reads that and it knows, statistically, that vanilla is the most popular ice cream flavour in the world (by 30%). Mango might not connect with everyone. Mango is a little unusual. So it ‘improves’ and the version says:<em> I am Kristina. I like ice cream.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ntfEPvOpaUVvIBHJ" /></figure><p>You probably don’t notice. Or you think the AI is doing exactly what you asked it to do. Some people believe that the way to get better results is to keep asking the AI to ‘refine’ or ‘polish’ the draft. So you do.</p><p>“Make this better.” “Make it more relatable.” “Help it reach more people.”</p><p>The AI looks at what’s there, sees ice cream, decides it needs more specificity, pulls from what it knows is most liked and most shared, and the sentence becomes: <em>I am Kristina. I like vanilla ice cream.</em> You send it out.</p><p>Now someone reads that. They meet you in person and they serve you vanilla, because that’s what you said you liked. You look at the vanilla and think, well, this person must like vanilla because that’s what they’ve given me. You eat it. You have a perfectly pleasant time. And neither of you ever knows that mango was on the table.</p><p>That is the Vanilla Effect.</p><h3>What I’m observing</h3><p>Across 50 CEOs and Chairs interviewed last year, it was clear there is no singular ‘flavour’ in leadership and that is important. The way a leader sees a problem, the cognitive style they bring to a room, the specific set of values that shapes how they make decisions under pressure, these are not soft edges to be smoothed out in the pursuit of broader appeal.</p><p>I started noticing the vanillaing before I had a name for it. Improvement that had worked exactly as designed, and in doing so had taken everything specific and made it general, everything distinctive and made it safe. ‘Change’ becomes ‘shift.’ ‘Love’ becomes ‘care.’ And a new group of words like ‘delve,’ ‘elevate,’ and ‘foster’ are everywhere. So much: “at the intersection of” or “no one is talking about.”</p><p>This is not an argument against AI. I build with AI. However, the “AI voice” is a result of pattern matching. When a language model predicts the next word, it often picks the most statistically likely professional word based on its trained data set.</p><p>Most AI writing tools are aligned to improvement on what performs, what the majority of people respond to from its data and bias. That is the right alignment for some things. But it is the wrong alignment for individual human identity.</p><p>The algorithm cannot tell the difference between a rough edge that needs smoothing and a distinctive edge that needs protecting. It does not know that mango is not a mistake. It does not know that the thing that makes your communication sound like yours rather than anyone else’s is precisely the thing your audience has learned to trust.</p><blockquote><em>It just knows that vanilla performs better in aggregate, and so it gently and invisibly moves everything toward vanilla.</em></blockquote><h3>The compounding problem</h3><p>When a leader loses their voice to the algorithm, they do not just lose their voice. They lose the capacity to bring something different into the room. Their team starts making decisions in response to a vanilla version of their leader’s thinking. Their organisation starts to reflect a blanded-down version of the values and perspective that made that leader worth following in the first place. Over time the whole system moves toward the mean.</p><p>Now scale that across an industry. Across every leader using AI daily to write, communicate, decide and present themselves. All of them being gently guided in the same direction by tools trained on the same data, optimising for the same outcomes. Blah!</p><p>The implications go deeper, as the algorithm is not trained on all human expressions. It is trained on the data that exists online, which represents roughly 30 percent of available human knowledge and experience, and that 30 percent is not evenly distributed. It skews heavily toward certain demographics, certain geographies, certain languages, certain ways of seeing the world. The vanilla it is optimising toward is not a neutral average.</p><blockquote><em>The people least represented in the training data are the ones being most aggressively moved away from their own flavour toward someone else’s.</em></blockquote><p>Then real life starts imitating the output. Leaders communicate in the language the AI taught them. Organisations start to reflect those patterns. That content gets created, published, shared, and fed back into the systems that learn from it. The bias compounds. The algorithm gets more confident that vanilla is correct because now there is more vanilla in the world, because it put it there.</p><p>And this is the part that keeps me up at night. We talk about AI starting to sound human. I think the real risk runs the other way. We are already starting to sound like AI, and some of us have not noticed. The words we reach for, the way we frame ideas, the instinct to smooth and generalise before we have even opened a tool. The training is not just in the model. It is in us now.</p><p>This is not a technical problem that will be solved in the next model release. It is a structural problem with how we have decided to define improvement, and it is happening at scale, right now, in every organisation that has traded its voice for a software shortcut (efficiency) and thinking to tools aligned to the mean.</p><h3>What the counter looks like</h3><p>Before you hand something to an AI to improve, ask: what is this system optimising for, and is that the same thing I am optimising for? If you want broader reach, AI optimised for engagement is doing exactly what you need. If you want to preserve your voice, you need something aligned differently.</p><blockquote><em>Something that knows what mango is and why it matters.</em></blockquote><p>Treat your distinctiveness as an asset worth protecting, which means being able to name it. What is actually specific to how you think?</p><p>And be genuinely sceptical of improvement as an uncritical goal. Generic is not the same as good. Popular is not the same as right. You are not trying to be everyone’s leader. You are trying to be a specific leader for specific people, and those people deserve the real version of you. Know your flavour. Protect it.</p><p>This writing was written using my own anti-vanilla prompt. It’s a set of instructions I put in front of any AI project to keep it sounding like me rather than like everyone else. If you want it, comment ‘No Vanilla’ below and I’ll send it through.</p><p>And if you want a space where that thinking is in the core logic of how an AI works, comment your favourite flavour. We’re opening Leadership Companion to the first 100 founding members. An AI advisor built around your voice, your business, your way of thinking.</p><p><a href="https://www.leadershipcompanion.co/?utm_source=linkedin&amp;utm_medium=organic&amp;utm_campaign=founders&amp;utm_content=kristinaLI">Leadership Companion — your thinking partner</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cd338c982887" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>