How AI is Going to Change Digital Agencies Forever
When I started Draft.dev in 2020, our value proposition was built on a “high-touch, high-quality” offer. It took four to six weeks of meticulous human effort to produce a single piece of technical content.
Today, that same process takes us two to three hours, and despite what you might think, the quality is indistinguishable to even a highly discerning technical audience.
While AI thought leaders (with obvious agendas to do so) are shouting about the “end of white collar work,” I’m seeing something more nuanced yet equally important on the ground: buyer behavior is shifting, prospects are arriving more informed, sales cycles are being compressed, and LLM-driven leads are starting to outpace traditional search.
It’s clear that generative AI is not just a handy new tool. We are facing a fundamental change in how digital work will happen and how products will be sold.
The choice for agency owners is no longer about if you’ll use generative AI or not, but whether you’ll try to hang on to the last vestiges of a dying model, or reinvent your business to thrive in an age of faster cycle times, better informed buyers, and higher expectations from service providers.
How the Ground is Shifting
I already mentioned the Draft.dev transformation, but let me elaborate a bit.
Using LLMs to help draft written content faster is probably the most obvious use case, and the tools are getting better at pulling in context from documentation, meeting notes, and previously written content. While we spend a lot of time gathering all this information from clients, once we have it, the writing process is significantly faster and requires much less editorial oversight than it did a couple years ago.
With a single senior engineer in the loop, we can typically create long-form content pieces in a few hours of work, whereas in the past, it took weeks of editorial and technical review before pieces were ready for clients. And, this new AI-powered workflow makes the output more consistent. We can incorporate previous comments from the client into our prompts, ensuring they never have to give the same feedback twice.
With this meaningful improvement in production cycle time and costs, we’ve been able to layer in content strategy, site/content audits, SEO/LLM monitoring, and distribution assistance without raising prices. Essentially, clients are getting more value for money from working with Draft.dev in 2026 than they ever have before.
But, I’m also not convinced that the tools will get to a point where they operate independently on high-stakes tasks anytime soon. LLMs and agentic tools are still just computers, and you can tell that they operate like computers, giving unreliable output when faced with novel input that’s not been previously tested or tried. So, each time we bring on a new client or attempt a different type of content, we have to re-tune our prompts and check the output more closely.
That said, any agency will recognize that having fewer humans in their COGS is likely a good thing for the bottom line. Fewer people per dollar of revenue means we don’t need as many managers, software tools to manage them, or communication overhead. It’s ultimately good for our businesses, but also good for clients as we can weather the inevitable budget ups and downs without as much swing in headcount.
Buyer Behavior is Changing
Another important consideration for agencies is that the way clients are finding and selecting agencies is changing. At both our companies (Draft.dev and The Podcast Consultant), LLM discovery is starting to outpace traditional search for buyers who book sales calls.
While SEO may still lead to site visits, traditional searchers tend to be higher in the funnel, and need to go through the full cycle of read => download asset => nurture => sale, while LLM traffic tends to book a sales call directly.
The reason for this is that buyers are using LLMs to dive deep into their specific needs and both our agencies are positioned to be specialists in their respective areas. So, when someone asks ChatGPT “what are some good developer marketing agencies?”, Draft.dev is much more likely to appear based on our internet-wide content footprint.
That said, these new buyers coming in from LLMs are better informed and often know more about what they want than people who come in from referrals or traditional search. I notice they tend to be evaluating us alongside competitors more often and they’re often more familiar with our workflows and service offerings.
The takeaway here is that for agencies, having published content about what you do and how you do it is no longer optional. The more specialized you are, the better and the more you write and talk about that specialty, the more likely you are for LLMs to recommend your services.
The Tech Stack of the Future
In addition to AI changing client services and discovery models, it’s empowering a suite of new software tools that will reduce operational complexity and enable smaller agencies to do more with fewer people.
A few examples of this include:
- Vibe Coding (Especially for MVPs): It’s never been easier for agency owners with small tech teams to build out a software tool for internal (or external) use. Our team uses Airtable, n8n, and a variety of LLMs to automate everything from onboarding to reporting, and we have essentially created our own in-house software to do it with very little cost or maintenance overhead.
- AI-Based Software Solutions: AI-powered PEO or payroll providers like Warp and bookkeeping tools like Docyt allow agencies to wait longer to make costly back-office hires, keeping OpEx low while you get off the ground.
- The Rise of Agentic Platforms: Most recently, agentic platforms like Zo.computer and OpenClaw have replaced my administrative assistant, and are now at a point of speed and reliability where I can trust them to handle a wide range of low-level, non-specialized tasks with a few minutes of skill-building.
All these factors will mean smaller, more nimble agencies with greater specialization and highly automated workflows will thrive in the next few years. While some clients will certainly use AI to build these processes in-house, plenty will have bureaucratic and technical reasons why they can’t adopt these tools quickly.
That’s where the big opportunity in digital agencies lies this year, and that’s why I see two paths for agencies going forward.
Two Paths for Agencies in 2026 and Beyond
The Legacy Maintenance Path
Some agencies will not want to change. Resistence from tenured employees, fear of change, and owners unaware of the impact this technology can have will try to maintain their headcount and highly customized client relationships.
The problem is that these agencies will be outpaced by up-and-comers if they don’t adapt.
It won’t happen overnight, but as more tech-forward, nimble agencies start building expertise, case studies, and a proven track record, they’ll chip away at slower-moving incumbents who can’t match their price, output, or capacity.
One of our jobs as agency owners is to standardize the work to a point where individuals can’t muck up a project as easily, and machines are an order of magnitude more consistent at performing similar tasks repeatedly than employees can.
So agencies that continue to rely on humans with high variability in skills will get left in the dust of those that take the other path.
The AI-Powered Agency
Agencies that can eek out more margin have a greater capacity to grow, stronger cash reserves, and the ability to bring on better talent. In short, they are more likely to succeed in the long-run.
So, it stands to reason that agencies that successfully replace headcount with AI tools will grow faster, be more prepared to weather a downturn, and start to win the hiring war.
The AI-Powered Agency will think like a software company (how can we automate this?) instead of a body shop (how can we train people to do this?), and while this has always been true, it used to be that software was less capable than people for certain nuanced, highly skilled tasks. With the right prompts and tools, generative AI is removing this limitation. You no longer have to employ a software engineer just to help you build a tool that transfers data from one API to another; you can just ask an AI agent to do it.
I no longer ask someone on my team to schedule meetings, make reminders to follow up with people, or research a client’s competitors. These things are all easily accomplished by AI, and with at least as much reliability as a junior to mid-level employee would have done in the past.
These tasks may seem trivial, but they stack up, and agencies that are best able to stack these wins will continue to thrive in 2026 and beyond.
Finally, the AI-powered agency will intuitively understand the value of showing up in LLMs today. As more agency evaluation moves to LLMs and is potentially done entirely by agents, this moat will continue to broaden.
Predictions and Tactical Advice
First, let me say that I don’t think the adoption curve is as short as tech leaders want you to think, but I do think it’s coming fairly soon.
The technology is moving fast and interfaces for AI agents and LLMs are improving, such that tech-savvy users can pick up these tools and be pretty effective with them in a few hours.
But, many industries are just inherently slow to adopt new tools. Regulated fields like healthcare and governments won’t change their vendors overnight, and legal and compliance departments in large companies will need time to figure out the implications of these tools.
That said, compared to other transformational technologies (railroads, electricity, or even the internet), the impact of generative AI on white collar work will happen much faster. The “deployment phase” is significantly shorter because of the groundwork laid by these preceeding technologies, and so much of our economy is wrapped up in things that AI can now do that the economic opportunity is immediate and significant.
My prediction is that 50% of agency jobs today will be done by AI within 2-3 years.
I think the jobs that will remain are client-facing ones (real-time video agents still can’t shake hands at a conference) and highly skilled or strategic client service roles.
Junior to mid-level roles that are largely focused on execution and back-office administrative roles will be much less necessary early on, and even for agencies that scale up and need some of these hires, they won’t need nearly as many as they did in the past.
Much like the “typing pools” of the 1940s and 50s, these masses of lower skilled jobs will just be gone in the next few years.
Key Takeaways
- For Employees: Don’t fight AI; become the orchestrator/advocate for it. The employees we have who will stick around are the ones who lean into AI to become 10x more effective than those who don’t.
- For Owners:
- Hire or appoint an AI implementer. You can be the visionary who sees how AI could be used, but an engineer or SME will help you get to successful implementation much faster.
- Increase expectations. Employees should have a higher client-to-employee ratio than in the past, but make sure you roll out the tools and processes to help them bridge the gap.
- Invest in AEO. As mentioned, this is becoming a huge part of the buying cycle. Ensure your agency shows up when the “buying agents” of the future do their research.
Leverage is the Only Path
Entrepreneurship is all about leverage.
While AI means change, and I understand that change can be scary, it’s also a huge opportunity to create time leverage.
Fewer people will be able to do more things. Some jobs won’t be necessary, but new jobs that were previously impractical will become possible. These shifts will cause huge changes in society but they’ll also provide new opportunities for savvy owners who are willing to lean in and make the most of them.
What do you think? Tell me how AI is changing your business by sending me a message on Linkedin.