Where agencies are right now
The stream began with Kirstie conducting a quick poll of the audience, which was primarily agency-oriented. Office Hours is, after all, part of our Agency Success program.
The poll results were telling: plenty of people use AI for content and copy, with a good number doing research and strategy. Fewer ventured into custom tools and automations, and only a handful had more unusual setups (building AI into client systems as a web developer or using AI to troubleshoot WordPress plugins).
To Nathan, the conclusion was straightforward: most people are still figuring out which of their everyday tasks AI can actually help with. The obvious entry point is content generation as it’s fast and visible. The bigger and more time-consuming shift is treating AI less like a smarter search engine and more like a tool (or even a teammate) that can execute processes you’ve already designed. As Nathan himself put it:
What does it look like to give it access to the tool I've been using, and what happens when I combine that with access to a new tool and start building these layers?
The chat interface is the starting line, not the destination
One major idea of the livestream was, and one worth remembering, that using AI purely through a chat window is, in Nathan’s own estimate, only a small slice of what those tools can do.
If you're just using AI as a chat buddy, basically, that's maybe 10%.
The real level-up comes from connecting AI to your files, tools, and existing workflows. However, the most important part here is to be willing to make the change. In Nathan’s agency’s case, they documented their entire delivery process: from onboarding to content architecture and even voice guidelines. It lives as a folder of markdown files a code-aware AI tool can reference directly, rather than the team having to explain the context every time.
Now, why does this matter? It all comes down to how generative AI works. Those tools predict the next word based on context, and the longer a conversation is, the more crowded the context gets, and things can fall through the gaps. Nathan, who’s tested many such tools, believes it can happen with all generative tools, even the best ones.
A documented process eliminates this “wandering off track” or at least makes it far less common. For agencies still living in the chat window, the practical takeaway here is to start with a low-stakes, repetitive task. It could be payroll calculations, sorting a folder of client-uploaded images, or drafting routine emails. Give your AI tool access to the files and systems involved (just not too much), so you don't have to copy and paste all the time.
It’s during such low-risk moments when things usually click and agencies realize the potential of AI tools.
Talking to clients about AI without overthinking it
Another useful topic was about client communication, partly because it’s an area where agencies tend to either over-explain or avoid altogether.
In Nathan’s experience, agencies don’t tend to itemize every tool they use as clients care about outcomes and cost, not whether you are running Photoshop or a particular form plugin. In contrast, AI does get disclosed, but it needn’t dominate every client conversation unless they ask.
Where it’s worth explaining, or over-explaining, is when a client pushes back. What looks like resistance is typically leftover opinions or objections from an earlier, rougher generation of AI output. In Nathan’s words:
I've not yet encountered a client who was philosophically against using AI.
The discussion with the client shouldn’t be around the morality and ethics of AI or AI in the abstract. Instead, it should be reassurance that someone is still reviewing everything that the AI produces, the same way they would with a junior team member.
Faster doesn’t always mean cheaper
During the livestream, this sentiment came up a few times in different forms, and it’s one that many agencies are quietly wrestling with. If AI speeds up a project that would otherwise take three months, should pricing reflect that?
Nathan and his team stopped connecting time and price years ago, as he admits himself, and it’s been something he’s been telling agency owners for years. In his opinion, value-based pricing was already the better model even before AI entered the picture. Now, it’s arguably better, and if you want even more information on the topic, watch our “Dialing in your pricing” livestream where Nathan discusses exactly this.
As to what’s changed, it’s not what agencies deliver. It’s how long it takes them to produce it. Whether that will lower prices across the industry, neither Nathan nor Kirstie was willing to call.
Keeping up without burning out
The livestream closed with something many agency owners struggle with: the sheer exhaustion of staying current. Nathan was absolutely candid:
Can we just acknowledge the fact that it's a full-time job to keep up with all the advancements in AI?
Even he admits doing it full-time is overwhelming. He has to run weekly AI news segments in Office Hours, hosting a mastermind group that talks about new developments, on top of everything else he does with AI for himself and his agency.
And while there’s no tidy fix here, to Nathan’s credit, he didn’t try to pretend there was one for the audience’s sake. There was, however, a common thread in the chat. Many agency owners shared that they treat AI exploration as an ongoing habit, rather than something with an end date. Some had a standing slot in a weekly team meeting; others had time reserved for tinkering outside of client work. The most vital advice some of them gave, though, was that accepting the feeling of being slightly behind is perfectly normal.
The key takeaway from this collaborative session
There was a thread that permeated this entire session: the agencies pulling ahead aren’t the one with the most AI tools. Instead, the ones turning AI use into a repeatable system rather than one-off wins are succeeding.
So, pick a tool, document the process, give your AI of choice access to real files and context, and build from there. You will see how everything else will follow naturally.
Finally, if you have any follow-up questions from this livestream or anything else you’d like to ask about AI, agency work, web hosting, or WordPress, join us for Office Hours. It’s every Thursday, at 2 p.m. EST, and Nathan and the community are all happy to help.
FAQ
How is using AI as a "teammate" different from using it as a chatbot?
A chatbot waits for you to type a new request every time. A teammate has standing access to your files, tools, or systems and can carry out a process you've already defined without having to re-explain it each time. The shift Nathan describes is less about the AI getting smarter and more about giving it a fixed job to do inside your existing workflow, rather than starting from a blank prompt every session.
Why does documenting a process help reduce AI mistakes?
Every AI model works from a limited context window, the information it can "see" at once. As a conversation grows, earlier details get pushed out, which is part of why mistakes creep in over long sessions. A written process the AI can reference each time acts as a fixed anchor, so it's working from the same complete information on attempt one and attempt fifty.
Generally no. The same way agencies don't typically itemize every plugin or piece of software they use. Most clients care about outcomes and cost, not the specific tools involved. AI should be named in contracts as standard practice, but a detailed breakdown is usually only necessary if a client specifically asks.
If AI lets my agency finish work faster, should I be charging less?
Not automatically. Speed is a function of your internal process. The value a client receives from the finished work hasn't changed. Pricing based on the outcome and expertise you provide, rather than hours logged, holds up well regardless of how quickly AI helps you get there.
