Designing for AI without losing the human with Ruth Hendry

From The Inside Seat Edition 2: A series of candid conversations with the community we’ve grown with for 18 years

In this transcript, Ruth Hendry, founder of Ruth & Co and longtime friend of Optimal, shares her perspective on how content design and content operations are changing, and why this matters for people who build products. I mean we hear it all the time: No AI without IA!

Taxonomy and content strategy are essential because AI needs structured data to function effectively. How content is organized, structured, labeled, and linked together directly affects the outputs we get, like the answers we get from LLMs, and therefore how much we can really trust them.

Ruth has a Wellington-based content operations and strategy consultancy working with mid-to-large organizations across Aotearoa and Australia. You may have heard her talk at UX New Zealand in 2023 on creating an IA strategy for diverse audience needs:

Looking at a business’ current IA is a window into their soul. It tells you what’s important to them, what they value, and the stories they want to tell their customers.

How are the ways we write and organize content changing?

We’re designing for humans and machines now which means you need to think about what your AI strategy needs from your content to succeed, like how can you get your content to show up accurately in LLMs like ChatGPT, what content does an AI-powered chatbot need for accurate answers, or what context do your customer service agents need to meet your customers’ needs. All these AI strategies rely on structured, accurate, consistent, and machine-readable content and audience expectations are really rapidly changing.

Information-seeking is the primary use case of AI now, and we can expect to see growing numbers of people using agents to fetch and act on information on their behalf. So an example of how this is affecting the way in which we write and organize content. I recently reviewed 8 government AI initiatives from across the world to understand what made them successful, and it was really clear here that the content itself is the most important indicator of success for those AI initiatives, and that’s because the crucial thing is structure.

So it’s effectively treating content much more like data, which is a really big shift in the way that organizations have previously thought about their content.

What are the key things to consider when creating content?

Increasing consistency and decreasing duplication.

Especially if you’re a business that produces a lot of content. If you’ve got content that’s duplicated across multiple web pages or press releases or out-of-date news articles and PDFs, and especially if that content is saying similar things, but in different ways, AI is really going to struggle to use it to give accurate answers. So, decreasing duplication, increasing consistency is really important.

Make your content and also your business explicit.

If you’ve got business rules, for example, like what help people might be entitled to from a government agency, or things like whether your airline allows children on flights unaccompanied, they’re the kind of business rules that we need to make really explicit and succinct and clear on our webpages and on our products now.

And this extends to things like your pricing.

So things like if your product’s got different pricing tiers, marking them up with a pricing schema, making sure you’re answering any questions that people might have about them. All of that is exposing the way that your business works to both people and machines, and that’s what helps people take action on that information.

Make it concrete and specific.

When documentation is structured well, it helps the rest of the business understand and embed these systems across the organization. In order to do that, you need to make things like your tone of voice guide really concrete and specific, so it’s not enough to say things like, ‘oh, we’re really human’, or ‘we’re really friendly’, or ‘we’re never formal’.

You need to give clear and specific examples of what that looks like, so we always say this, but not that. You’re exposing and making explicit all the content choices you’re making, so that an organization can have a sensible discussion about them.

Are there differences in optimizing content for humans and optimizing for machines?

Where there’s alignment between optimizing for humans and optimizing for machines is that content always has to answer a real need or question. Increasingly, we’re all able to ask more personalized or niche or unique questions, so that long tail of search is really increasing. But we still need to focus on what those people want, no matter how niche and specific those questions actually are.

There’s a convergence because we care about clarity, scannability, plain language, and logical structure too.

If you’ve got a well-structured page that answers a question, and it gives somebody what they need, and it’s got clear logical heading hierarchies, and front-loaded key information that serves humans, screen readers, search crawlers, agents, everyone benefits from that.

The conflict is that sometimes machine-readable content demands a kind of redundancy and explicitness that human-facing writing avoids. Ruth uses the example of looking for a home loan, when you know you’re on your existing bank’s website. You don’t need it to spell out your first home loan in every single paragraph, but machines do often need that to be spelled out.

I think the answer to balancing those tensions is probably in using structured content. So that’s where you separate the information from the presentation layer, and you use markup, preferably with a good content model that sits behind it. That’s the way you can serve both the human and the machine audiences without compromising either. So it means, in that example, you can write about your home loan in a way that makes sense to people, and it appeals to them, because obviously we need to know what a product is, and we need to want to purchase it. There’s an emotional component there. And then you can layer on the machine-readable structure, so metadata, schema, defined relationships, and that comes through in the technology.

There’s also the consideration around optimizing content for findability and optimizing content for human enjoyment. You know, putting really obnoxious keywords that make it findable, improving SEO or AEO, but are dry to read and come at the expense of brand enjoyment.

You’re going to generate a lot of traffic, but the traffic won’t be doing what you want it to. Humans are still the ones setting the bar for what good is, and judging what good is as well, even when an AI agent is completing tasks on their behalf. So, balancing machine readability with content that maybe earns trust, or it reflects a brand, or it enables somebody to get something done. That’s exactly the kind of big picture thinking that content strategists and content designers are made for.

Content governance and cross-functional collaboration matters now more than ever.

Content governance is increasingly important, because that’s what makes sure your content sounds the same, it’s structured the same, it’s consistent, you’ve got a clear view of who your audience is, when you produce content, why you produce it.

Decisions need to be consistent across all of your content, no matter where the channel is. Picture you’ve got three teams, and they each own a different channel, and their content overlaps, and there’s no single source of truth. If you’re trying to build an AI chatbot on top of that, it’s going to give you contradictory answers, or if you’re trying to get your content to surface well in something like ChatGPT, it’s going to really struggle if that information source is inconsistent.

And you also need to have people in your governance groups who understand the implications of the decisions they’re making. Someone needs to have a great understanding of public-facing, multi-channel content publishing to understand digital best practices and how AI tools are moving. Somebody needs to be in there who is across the metrics and the data. Another who can advocate for the customer and see how customers are behaving, if there are increased calls to the call center, where people are dropping off without subscribing, or if they’re unsubscribing, for example. If you’re in a highly regulated industry like government or healthcare, you also need someone with subject matter expertise because you need to make sure that you’re not opening yourself up to any risk. There are many diverse perspectives with unique expertise which need to be represented in a room to launch successful products.

What matters most for content designers in 2026?

Think about how you describe your own work.

The first would be reshaping how either you think about or how you describe your own work. So, I imagine lots of people in the content discipline are already thinking like this, but I have switched more to describing my work as content as infrastructure, what I mean by this is that it is really foundational. It’s what everything else is built on. It’s expensive, it’s disruptive to fix when it goes wrong.

Hold the whole experience in your head: the emotional journey, the information architecture, the weird edge cases that only your business has.

The work that I think is really stubbornly and beautifully human is the judgement work. So that’s things like understanding what customers actually need when they can’t even articulate it themselves, or being able to clearly reason about when a content system for example, maybe it’s producing technically correct, but actually quite unhelpful outputs, making the call on when plain language is so stripped back that actually you’ve lost meaning and accuracy, because it’s got too plain. That kind of stuff requires holding the whole experience in your head simultaneously. That’s really, systems thinking across a business, and systems thinking about the people that you’re serving, and that is very hard to automate.

Understand which parts of the process are automatable and how AI can be implemented into the content discipline in novel ways.

Basic copy, microcopy, search optimization, AI can handle this pretty well now…Who likes reviewing copy for sentence case for the 50th time? Or auditing a whole website for really out-of-date terms, especially if you’ve got a website that’s thousands of pages long. They are the perfect jobs for a machine. So working with engineers, understanding the distinction between which parts of your process are automatable, which parts of your process aren’t, and the logic behind how and why you make choices to automate things. Or also where to bring AI into the content discipline in novel ways. That’s what brings real value to your team.

Everything is changing.

It can feel a little bit overwhelming at times, and a little bit challenging. I think what I’d really like to leave people with is maybe a sense of possibility or excitement, because I think this moment, despite the fact that I know there’s a lot of uncertainty, it is genuinely quite exciting for people working in content.

This article features Ruth Hendry, Founder, Ruth & Co: content operations, strategy and content design that digital experiences and AI ambitions are built on. Find her at ruthandco.nz and on LinkedIn.

Follow our LinkedIn to watch the conversation roll out, and follow for more candid conversations on how people are navigating shifts in the industry.

ICYMI, Last month we caught up with Clara Kliman-Silver on the UX ecosystem in the age of AI.