Character AI policy
Unless you’ve been frozen in ice for the past few years (lucky you!), you’ve seen how bullish a lot of the design and technology world are about AI.
I’ve worked with AI-driven companies and with companies who are using AI, sometimes in interesting and fruitful ways. Machine learning can be incredibly powerful when applied to datasets and processes that can’t reasonably be addressed by human labor. In content and UX, much of the AI discussion is usually around the use of large language models (LLMs) like ChatGPT.
The work I do with my clients doesn’t tend to lend itself to these types of tools, but a policy around LLM use can help us set expectations.
In short, I don’t specifically seek to use any tool labeled as “AI,” and I won’t recommend using an LLM unless it’s the best decision for you and your team and supports all other goals and working principles.
Here’s how Character currently approaches LLM tools, and how this relates to broader principles I apply with my clients.
This policy was last updated on July 19, 2024.
We won’t put your content into an LLM
Apart from standard autocorrect, or maybe implementing a basic tool like Grammarly, I generally avoid LLMs in my daily work. I’ve used them to help brainstorm things like taglines or headlines, but I would never copy and paste my writing or any company’s material into one of these tools.
There is too little transparency around where that data goes and what happens to it for me to take that risk. From an efficiency perspective, I also advise against using LLMs for drafting content, since it often takes longer to rewrite something that looks polished (but often makes no sense) than to turn messy notes and thoughts into something coherent. I promise that your notes and typos don’t scare me.
I believe in “minimum necessary technology”
One of my working principles is something I call minimum necessary technology. It’s part of a governance mindset, where I don’t believe in adding things without a plan for their maintenance and (perhaps) eventual retirement. When we’re working together, we’ll use your existing tools to their limits before I suggest adding a new tool to the workflow. Besides, every new tool is a potential security risk, another cost, another thing to learn and adapt to.
Our work will involve change and transformation, and I’d rather we focus on defining a new normal and working toward mindset changes than spend our precious time figuring out some SaaS product with a silly name, or get everyone onboarded to our shiny new approach to automatically generating (usually mediocre) content.
IA (information architecture) before AI
Machine learning has many useful applications, and if, in our work together, we find something that is worth automating, we’ll certainly explore whether it’s a worthwhile strategic decision for you in the medium and long term.
For example, if we do good work tagging content and that enables something like autosuggest for search or content recommendations, we’ll investigate that. In my experience, once things are organized and categorized properly, you might not need anything very fancy on top of it. But we don’t have to call it AI.
Sustainability comes first
In my work, I focus on outcomes and goals, not outputs and empty metrics. I like to help my clients make smarter, more informed decisions more reliably and consistently.
Most organizations need help making sense of things, not speeding up or doing more for the sake of more. I won’t knowingly encourage unsustainable behaviors, whether that’s moving you away from your own climate principles, or creating content for the sake of it. Quality over quantity, always.
If you’re using these tools already, we’ll discuss it
I’m coming in as an outside consultant, which means I don’t always have access to—or need—all of the software you’re paying for. If you’re already using some type of LLM in a workflow that’s relevant for us, we’ll discuss whether I personally need access to it.
It can be helpful to use all the tools you use, but it’s usually most important to have access to Figma, digital whiteboards, messaging apps, documentation or project management tools, and documents.