Episode 51 - Dock: Eric Doty on automation, learning new skills and LinkedIn superstardom
Read more
%20(3).png)
Katie is a Staff Writer at Every, where she runs Working Overtime—a column on how technology is rewriting the rules of work. She also contributes to a few other AI-related columns there (check out her articles here). She's also a big part of Every's AI editorial operations, exploring how to use AI to help the other writers at Every do their best work.
She also works with founders, investors, and teams to develop thought leadership content and build AI-driven editorial systems—stuff like style guides, author personas, and publishing workflows. Before Every, she also worked for Animalz, Startups.com, and a few other startups.
TL;DR: Katie summed up her own advice in three takeaways:
Context is king. Strategy docs, style guides, and any other kind of documentation are your friend, both for getting your current AI to behave the way you want it to and for getting new AI up to speed. Tend to your context like a garden and it will reward you.
Make AI do the work for you. If something seems overwhelming or scary or too technical, just ask AI if it can do it for you. Worst case scenario is it says "I can't help with that" but usually it can, even if it's just guidance (these days it will generally do the thing for you).
Practitioner-led content, always. The more content AI can synthesize from what's already out there, the more content differentiation will depend on contributing net-new knowledge and insights, which starts with people doing the work and sharing what works!
TL;DR: Katie is a big part of Every’s AI editorial operations, managing her own writing and using AI to help other writers and SMEs create their best work at Every, too. If you’re thinking about how to create great content and manage a more editorial / narrative voice for your brand or team, Katie’s biggest takeaways:
Every has one of the most distinct editorial voices in tech media right now. How do you protect that voice as you scale up content ops with AI in the mix?
A lot of it comes down to documentation and what I guess you could call "context engineering" — developing really robust resources that capture our style, not just in terms of grammar and mechanics (although we have skills for that, too), but in terms of purpose, intent, voice. Each column has a dedicated style guide (here's mine for Working Overtime, which I have tended overtime like a bonsai tree) that captures that high-level guidance, and we also have a publication-wide guide that sets the context around the entire company, what we do and why, who our target audience is, etc.
Another thing I'll highlight is that humans are VERY much in the mix. We've got two staff writers (myself and one other), plus our editor in chief, managing editor, and senior editor, so everything gets at least three sets of human eyes on it before it goes out the door, and the human really owns the "voice-iness" of the piece.
I've written before about the "timid scribe problem” with AI and how it tends to pull everything toward a safe mean. I see my job and the job of our other writers and editors to be "roughing up" those smooth edges and making sure everything reads authentically and humanly while also staying true to the brand.
What suggestions do you have for a brand new enterprise reporter within the nation's largest online university? How would you go about finding good stories/features? What would you tell someone just starting out in the journalistic writing world?
One thing that has really helped me at Every as we have started "reporting" on the goings on inside the company is that we have a culture of transparency. Conversations happen out in public, in Slack, which makes Slack a goldmine for finding interesting takes, new tools to explore, and other "nuggets" that can potentially get evolved into content.
So the biggest advice I think I would give someone who's just starting out and trying to treat their workplace like a “beat,” so to speak, is to learn the arteries that information flows through and position yourself so that you can tap into those arteries and mine them for content.
I'd love to get your thoughts on how important video is in a content strategy. Any data or use cases you’re seeing?
I don't have concrete data on our outcomes, but I can tell you that Every is very committed to expanding into video — we've started doing live streams to accompany our vibe check reviews of new models, we record and send out recordings of all of our camps (what we call webinars), we're experimenting with shortform, etc. We actually just put out a job description for a talent producer to help with all of this stuff.
I think putting actual people in front of the camera really is non-negotiable, especially with AI content taking up more of the real estate on the internet. People want to be able to see and connect with personalities.
The other bonus of video content is that it can be really fertile ground to mine for additional content ideas, whether it's follow-up videos, or a blog post to expand or codify something we said on a livestream. So video content and written content can form this virtuous cycle, which is something I've seen pay off at Every quite a bit.
Not all content formats reward “being human” equally. Where have you found that voice and weirdness actually create leverage and where is it basically wasted effort?
Voice is absolutely essential in most of our longer-form stuff, particularly when it's someone like Kieran, our compound engineering expert, writing about his newest engineering workflow, or me writing about how I'm getting AI to help me not fail at my OKRs.
Stuff that's more transactional, like feature updates on our products or upcoming event announcements, tend to be the kind of thing that are much easier to put on rails. We have skills set up to produce that type of content and I seldom feel the need to touch the outputs that the AI comes back with for those types of content.
What are your KPIs being editorial led?
We actually just switched these up! Historically we looked at open rate and page views, but these days we're much more interested in clicks: is the content driving people to take action? Whether that's clicking on a guide or a link to our github to download a skill, or just reading more content, we want to see that our content is being used and that people are making progress with their goals. Clicks are the best proxy we've found for measuring that in a data-driven way.
TL;DR: Every has a reputation for being one of the places for editorial content on working and building in an AI-first world. Katie’s main takeaways on how they’re managing this:
More on exactly how Katie’s doing this:
My engineering friends specifically cite Every as one of the few outlets they actually trust on AI. What does Every do differently in how it vets or produces that content for that audience? Because most companies get completely ignored by them.
The big thing here is that everything Every produces is grounded in first-hand experience. Since you mention engineering, I'll highlight Kieran Klaassen as an example: Kieran has pioneered a workflow called compound engineering that has gained quite a bit of traction in the community, And the thing about it is, it's 100% grounded in how Kieran actually works with AI day-to-day, and the workflows he's developed around the product he's building (Cora, our AI-powered email assistant). So it's not just theory — it's rooted in personal experience with receipts to match.
I think we've all had the headache of trying to produce content for an audience that we don't really know or only half know. And it's easy to say "leverage your internal SMEs" as some handwavy thing that will solve all problems. But Every is the first place I've worked that I feel like really nails that — everyone has their hands on the tools, everyone is learning how to work with them, and the content we produce is largely us "thinking in public" and it turns out that more often than not, if we're thinking about it, our audience is thinking about it too.
How do you decide what you're going to write about for Every? What's your thinking process/how do you vet exactly what you want to write about and what the narrative should be?
9 times out of 10 for Working Overtime I'm just writing about the most recent thing I've learned, what I've found annoying or useful, how a new AI tool has changed how I work or how I think about work.
The thinking is that if I'm feeling a point of friction, like compulsion to work more with AI, or struggling to multi-task with AI working on projects on all sides, chances are good that someone out there can identify with that and seeing someone think through it will be useful.
At this point, I've got the formula pretty hammered out: start with a moment of friction or tension, expand out to the larger issue that my experience represents, then unpack the practical steps or lessons I've learned from working through the problem, and land on something actionable the reader can take away.
That's a pattern that AI picked up in my writing about a year ago that I really like and have sort of codified as my "signature move" (although now I’m worried that it's getting too regimented and I need to break out a little).
How do you approach working with your team to understand the interesting things they’re working on, encourage them to share their learnings, and navigate through the potential noise to find the stories worth telling?
I've got spies (AI ones, of course) all over our Slack that are always on the prowl for potential content ideas. They have context like our Q2 editorial strategy, style guides for particular content formats, and a few other guiding principles, and they've been pretty reliable at surfacing the kinds of things we want to write about out of the product channels, the every-one channel where we share news and hot takes, etc.
The whole team also has their antennas out and will occasionally tag the editorial team and say "this is a nugget" (nugget is our term for an idea that's worth sharing, developing, etc.)
Once we're in full-on idea development mode, it's more of a traditional approach: if, say, Kieran wants to write a more longform piece, I'll work with him to shape it, whether that's hopping on a call to talk through or sending questions async for him to monologue about while he's on a walk, and then we tag-team development from there.
TL;DR: Every is an AI company, and Katie’s AI workflows and systems are *chef’s kiss.* Also, Katie works with founders, investors, and teams to develop thought leadership content and build AI-driven editorial systems. If you’re trying to figure out how to use AI for content in ways that actually work, Katie’s big takeaways:
Plus, Katie’s favorite workflows and Skills:
I’m building a lot of new stuff in Claude Code and Lovable right now. Building and perfecting these processes takes a decent amount of time.
One thing that worries me with the pace of AI development is that after sinking tons of time into building out some operation with AI, there will be new tech that is better and I’ll need to rethink the whole process within months. I envision this never-ending cycle of building new workflows.
At Every, y’all are constantly adapting your workflows to the new tools that come out. Does my concern resonate with you or do you have a different way of thinking about it?
Oh I think about this CONSTANTLY. Like, I was just confident that I was fully settled into the Claude ecosystem and then GPT 5.5 came out and now I'm yeeting myself back into OpenAI territory. Our CEO, Dan Shipper, has told pretty much all of our engineers to plan on throwing their code out every six months.
I think the good news is that, with each new evolution of the tech, more and more of the process can be managed for you. So whenever I'm adapting something new or trying a new tool, I start by asking the tool itself some version of "how can you help me? What information would you need to be most effective?" and most of the time, I have some existing documentation I built, whether it's a context document or a spec that I cobbled together (also with help from AI) and I can provide that context to the AI and it goes off and does all the work for me.
In developing new tools and plugins for subscribers, does your team consider how/whether they can be used in the enterprise (given security/privacy/regulatory concerns)? Do they worry that functionality will be included in the next frontier model, making the tool obsolete? Just curious what factors builders and superusers of AI consider, or don't.
I confess I'm not 100% up-to-speed on the enterprise of it all, although I know it's something that our consulting team thinks a lot about as they work with larger firms on AI adoption throughout the org and keep running into red tape lol.
I think if we spent too much time dwelling on whether AI is going to eat a certain capability that we're building around, we would never ship anything. It's just a reality of the way the space moves now and we've come to terms with that. Every time a new model releases and some paradigm shifts in terms of its capabilities, we just realized that we were going to have to build around that new possibility.
What advice do you have for enabling other writers with AI? As in, how do you make your AI systems more multiplayer? How do you scale that across a team?
Oh man, this is a question that haunts my nightmares. I don't think that we'll ever get to a point where AI is truly multiplayer in the sense of everyone interacting with the tools in the exact same way. I think everyone will have their own personal workflows just the same way that everyone has their own workflows for "trad writing", so to speak.
That said, I do think there are certain paradigms and workflows that help. I'm really bullish on skills as a paradigm. I've spent a lot of time lately codifying certain content types that we produce regularly at Every into skills that ensure that that kind of content gets executed consistently no matter who's actually manning the writing.
And skills are something that can be shared throughout a team space, if you have, for example, a shared Claude workspace. So that's an example of making the same set of context and skills and instructions available so that everyone has access to them and making the experience more multiplayer.
(Shameless plug: Spiral is evolving into more of a multiplayer system where you can share knowledge/context and prompts and styles for access across a team. So that's an example of solving for the multiplayer challenge. But I don't think it's a solved problem yet, by any means).
I'm hearing that the marketing teams of the future will be fluent in Github, and run on Github. After digging into it, I see the rationale and think it is a great tool, but just not sure it's realistic to expect most marketers to start thinking in commits and repos instead of slide decks and folders. Do you find the switch to an AI tech stack has been easy at Every? How have you managed it?
It wasn't easy at first, but it's getting easier as the tools get more sophisticated. For example, I've found that desktop coding tools like Claude and Codex make the process of interacting with Github repos a lot more intuitive.
In general, I spend a lot of time telling AI "I need to [insert task here]" and letting it work out the rest. With advanced models like Opus 4.6/4.7 and GPT-5.5, it really can handle most of the in-the-weeds stuff and let me focus more on the outcome I'm trying to drive and whether what the AI gives back to me meets my needs or not.
I'll also say I don't think we've achieved the final form factor, particularly for versions of work that are NOT code. Because coding was the first killer use case, the industry is making everything code-shaped, but I think the free market will free market and as demand in areas like marketing becomes clearer, dedicated solutions will start popping up.
If you had to build your own road map to get a content marketer caught-up on AI best practices, how would you do it?
I would start with my absolute favorite AI workflow, which is opening up a chat, telling it the objective that I'm working toward, and having it interview me to extract the information it would need to successfully execute on the task.
So I would start with something like "I want to produce a guide for content marketers, helping them get up to speed on the latest AI capabilities and how to incorporate them into their work. Interview me one question at a time to extract what I think on the topic" and then answer the AI's questions, throw it some context like articles I consider to be canonical texts on the topic, skills and other workflows that the reader can just grab and steal, things like that. And then I would shape the content from there.
What are your most used (or most impactful) AI workflows/skills at the moment?
My most impactful workflow at the moment is probably the writing agent that I've set up. It goes through the full writing pipeline from brainstorming and ideation to outlining to drafting to editing to finalizing, so I can go to it with something like, "I have an idea for an article about my experience creating an AI project manager to manage my OKRs," and it will pick up the idea, interview me according to what it knows about what constitutes a good Working Overtime essay, and then we're off to the races from there.
For more from Katie:
.svg)
%20(2).png)
