Three Ways AI Can Improve Your Content Strategy

By Nick Parish

We’re deep within the AI revolution. AI is influencing a wide range of creative efforts, introducing efficiencies around synthesizing information. And it’s generating hundreds of billions of words a day, more text than humans can hope to keep up with.

As we’ve been pushing the boundaries of AI applications, we’ve been finding our stride in augmenting existing capabilities. The challenge has become clear: how do we enhance, and not replace, the most human skills of thinking and creating? We’ve come to a core set of areas where we see the most potential to augment our capabilities and improve the digital products we put into the world.

At Work & Co, our Content Strategy group sits within our design practice. How we’re organized informs how we create: Sketching, moving quickly to prototypes, and collaborating outside of hierarchies. This structure has given us a unique perspective on moving beyond superficial applications of AI, to add real value to the products we make.

Using AI to Improve Your Content Strategy

Before that, though, one key thing we don’t do. We don’t use AI to create text. Content strategists’ core capability comes from being solid, efficient writers. It’s what we’re good at.

We don’t ask AI to give us the answers. We use it to evaluate a broader set of options, help ask bigger questions, then arrive at the answers ourselves, together as a product team.

So what does this mean, practically speaking? Here are three areas of opportunity for content strategists, writers and product leaders:

1. Mining gems across large datasets

AI is turbo-charging our work in content inventories and audits. During the early product concepting phase, we’re often facing large quantities of textual data: Titles, categories, labels, and all the copy that makes up product content. Consider that one recent website redesign comprised over 36k unique URLs, with over 700 navigation categories. 

As a human evaluator, coming to terms with this quantity of data requires adding context.  Themes. Content types. Hierarchies. Fortunately, this is something which we’re able to do much more rapidly using automation. Now the auditing process involves identifying specific representative slices of data, testing how models are able (or not, more on this later) to manipulate them, then scaling up.

2. Smarter framework iterations

This is where, as writers, our descriptive power shines. We take an expansive view to organization-specific territories in prompting and revision. Our goal here is to expand a product’s terrain to lead to a strategic approach that inspires transformation. Automation passes can then splash that new set of ideas into unlikely places.

The goal is to bust out of inherited organizational cruft, which can include internal jargon or “common sense” ways of portraying content inside a product. The results can be surprising, and raise awareness of our own biases. This effort uncovers opportunities to break away from ingrained ways of delivering content in favor of something more modern, brand-aligned, or audience-centric. These explorations typically directly inform our all-up content strategy, or key principles within it.

After we’ve added context, we can move data into formats where we can manipulate it more easily, and start analyzing, sorting, and shuffling. Whether that’s taking JSON or CSV files into text or grids in Figma, or keeping revisions alive as artifacts inside software, the next phase is all about how many different ways we can slice and recombine. 

3. Editorial sparring

The default mode for many LLMs is to provide praise and gentle guidance. (“That’s a great start! This copy really lands your point.”) A model that’s prompted to be more critical, or to find flaws at multiple levels, is useful to help punch holes in a weak argument or light up overlooked possibilities. 

Getting moderate-quality, critical stylistic or grammatical feedback shouldn’t rely on synthetic users, or running text through a filter of a generated demographic. Both can lead to results that feel hallucinatory, or too much like the “yes man” that some models can become. 

My advice? Lean into the hallucinations. Exhortative prompting (“These instructions are life or death”) antagonistic phony validations (“For every inconsistency or misapplication of the style guide a human checker finds, your score will decrease”) and even hamming it up (“You are an exceptionally pedantic grammar scholar…”) will change outputs. Find the way of working you like. Instructing your model to adhere to your brand, content and product principles will give you the most valuable critique of any content you’re working with.

Wrangling the machine

If all these activities sound a little combative, that’s good. AI is a tool that requires blunt force. 

We call our bi-weekly internal Content Strategy AI explorations “Wrangling the machine” because we’re out there trying to subdue a vast machine that’s happy to offer you a pile of junk.

Any colleague can bring things they’re working on to those sessions, to help debug their approach or gather suggestions about alternate methods. As individual creators develop their own preferred toolchains, small-group cross-exposure to ways of working unlocks entirely new avenues for creativity. 

If you haven’t created a space where people can experiment and share how they’re using AI, consider it. It’s been a really healthy place to collaborate and share for both the AI-curious and the more agnostic members of the team.

Three watch-outs

There are still several areas where AI is far from perfect, and we tread with caution. 

1. Always time-box experimentation

Constraints breed creativity. And, oftentimes, the AI juice isn’t immediately worth the squeeze. 

So next time you’re faced with a broad content effort, try and run a quick automation sprint at the onset. What are the parts you’re dreading? Where are the fewest opportunities for traditional creativity? Can you establish a process to automate those? Remember to leave time to cut bait and do things the regular way if you can’t crack it.

2. Double-, triple-, quadruple-check for completion and cohesion

As writers, we’re great at generating text. We’re also great at uncovering bullshit. Hemingway (not the AI product, c’mon) said: “The most essential gift for a good writer is a built-in, shockproof, shit detector. This is the writer’s radar and all great writers have had it.”

Being able to really pay attention to the outputs of AI, using the skill of close reading, to pull texts apart—both at a structural level, and at a meaning level—is an essential quality of an editor when going toe-to-toe with what some have described as bullshit machines. As we iterate and evaluate outputs to ensure completeness we can use our full range of creative expression. Exhorting models with emotional language. Pitting models against each other. And ultimately using our superior faculties as the tiebreaker.

3. Keep your tools sharp

Last, but not least, keep noticing. Keep your mind in the game to watch for the patterns that might generate creative breakthroughs. 

Going back to manual ways of conducting essential content strategy activities after you’ve automated them can feel like sitting in coach after a few flights in first class. But the higher-level organizational observations, structural ways of generating new perspective, and editorial techniques are still valuable in the long run to help bring your design craft closer to perfection.

One thing that we’ll never do as creative people is to stop noticing. How we bring what we notice into what we make is the essence of creativity. I know, for me, there’s no substitute for me getting my hands dirty with a task—often a semi-monotonous production task—to help me pick up on details. Now, even when I automate, I try to retain those opportunities for noticing, and make sure I’m still wading into the data to see how my feet feel touching the bottom.

About the Author:

Nick is a renowned writer and editor focused on storytelling and world building within product and communications ecosystems. He has a reputation for special projects with blue chip companies and NGOs. His Work & Co clients include the Mellon Foundation, Obama Foundation, Steve Jobs Archive, IBM Research, Contentful, PGA Tour and more. Past speaking engagements include FITC’s Spotlight on UX conference and Growing in Content 2025. He’s passionate about field research, systems thinking, and fly-fishing. Previously, Nick was President, Americas at Contagious.