
Intro: I wrote this article with Grok, Claude and Gemini reviewing the codebase of Agency. I asked the question, “Is this too powerful a system in one person’s hands?” because I think Agentic Software is about to rip a hole through some industries. I personally think the Agentic Age started for real in December 2025, with the launch of Google Antigravity.
Yes, I used Claude before that, and ChatGPT before that, but for me, it was Antigravity with that clear combo and ease of multi-intelligence switching. It appears I was on a similar but different track from Anthropic, too (I would hardly know; I feel as if I have been down a mine for almost a year). I mean, Claude is great. They employed skills properly before anyone else. I created and published what I called Agentica – and wrote a paper on skills to improve LLMs too (frankly I wasn’t sure if I was being fantastical at the time, but there were indicators) – a few weeks before Claude announced skills, but it was way back in 2023 when I first identified that these LLMs were going nowhere fast without “plugins” – or skills.
And that’s what happened.
But instead of opting for open claw (too anarchic) or Claude skills (designed to make you a great, almost Iron Man type EMPLOYEE) I embarked on my own mission. To build a tool, using Agentica, that can actually help a user achieve 10X high-quality output. I mean, that’s the aim here. I could have built a slop machine much faster and easier than this space station.
The truth is, I wonder if these “skills” as they are, are a race to the bottom. At the end of this process, you have a workforce of highly techncial, super-powered employees all looking for a job. And I’m sure Claude is training on every one of those skills and one-shotting most people’s apps with every update. I mean its almost as if we, through the open source community, are training systems to replace us and our skills.
I’ve published as much as anyone in the SEO space over nearly 3 decades.
It might be time to lock our skills and fresh experience away from the global AI setting out to replace you.
There are a lot of different facets to the agentic web, and it is going to need guardrails. Even for me. I’ve built Google’s helpful content guidelines into the very DNA of this thing. There is no back door. Even I, as a super admin cant get around it. I am using the same system as you.
Strategically, I’ve burned the boats in that direction. There is only one way forward.
Help users create content and art that’s not AI slop. Or what is the point? Another thing I’m committing to. No impersonating humans on third-party sites. We need to get AI out of our spaces and get it working for us in specific spaces we own.
The point is to get AI working for us, not the other way about?
I digress, though, that’s for a later post, but I wanted to be clear the direction I am going with this.. agentic operating system.
Let’s proceed from the original position of the article.
Is Agency, and agentic systems like it, too powerful a system in one person’s hands?
There is a fine line between a productivity tool and a weapon of mass disruption. After reviewing the architecture and capabilities of the Codename “Agency” platform, it becomes clear that this system is walking that line.
As the platform’s own documentation states: “Agency is not just software. It is a completely simulated, autonomous digital marketing agency. Agency has bridged the gap between raw intelligence and human-in-the-loop operational execution.” Built to replace an entire digital marketing operation, Agency doesn’t just assist a user; it multiplies them.
Here is why a platform of this magnitude represents a fundamental shift in how single operators wield digital influence.
1. The Death of the Corporate Hierarchy
Traditionally, scaling a business to excel in a niche required capital, human resources, management layers, and time. Agency completely subverts this requirement. With an orchestrated swarm of 80+ specialised agents, a single user can deploy role-based operational layers instantly. From research interns to compliance officers, roles are strictly defined, task delegation is automated via custom systems, and specialised expertise is routed dynamically.
One person now has the operational throughput of an entire mid-sized corporation, entirely bypassing the friction of hiring, payroll, and human management. Operating from a central “Command Centre,” the user acts as the CEO, viewing a dashboard where they see everything, approve everything, and direct the workflow.
2. Directed Execution and the Omni-Channel Vision
The “Agency” doesn’t just generate text; its architecture is designed to interface with reality through specialised, role-based assistants. While current public demonstrations heavily validate its high-speed SEO/AEO auditing and content workflows, the platform’s broader aspirational vision encompasses full omni-channel execution:
-
The Voice of the Agency: Elsa, Agency Company Secretary, acts as the first point of contact, dispatching tasks to office specialists and speaking to the user by voice directly from the browser.
-
Triage & Logistics: Margaret, the AI receptionist, is designed to manage multiple inboxes autonomously, while Michelle handles travel itineraries and Stephen monitors cash flow.
-
Rapid Deployment: The system actively excels at high-speed technical SEO audits, content gap analysis, and Quality Rater simulations—tasks that previously required days of chained research prompts and fact-checking.
In the hands of one person, this architecture represents the ability to project the illusion of a globally active enterprise, allowing a sole operator to maintain a massive operational footprint.
3. The Empirical Grounding Engine
One of the historical weaknesses of AI has been hallucination. Agency solves this with a brutal, 6-layer grounding architecture spearheaded by an internal “Research Team” of data specialists and fact-checkers. By ingesting live crawled HTML, Google Search Console data, GA4 metrics, and utilising deep query-based extraction, these AI agents do not guess—they analyse hard empirical data.
Coupled with the Quality Rater Engine, a single user wields a powerful tool for aligning with search realities at scale. This engine displays real-time SEO audit signals based on official guidelines, measuring metrics like Q Site Quality, YMYL Audit, Content Effort, and Topicality. By automating these checks, the system ensures outputs are rooted in measurable reality rather than generative guesswork.
4. The Amplification of Influence
When a multi-LLM dispatcher dynamically routes tasks to the best-suited model (Gemini, OpenAI, Anthropic) while a “Topical Experience System” ensures every agent actually learns over time—building real domain expertise through compound growth—the system’s leverage becomes exponential.
If one person controls the narrative generation, the data analysis, the customer interaction, and the public relations distribution, they hold a disproportionate amount of influence. The “Agency” platform isn’t just software; it is an industrialised synthetic workflow.
5. The Compliance Layer, E-E-A-T, and Combating “The Sloppening”
As AI tools become ubiquitous, the internet is facing what many call “the sloppening”—a tidal wave of low-effort, mass-produced synthetic content designed purely to manipulate algorithms. However, Agency’s internal architecture reveals it is specifically designed not to add to this digital pollution.
Because the ultimate measure of success is maintaining a genuinely customer-first approach, the platform enforces strict quality control. Agency addresses this through a multi-tiered Compliance Layer and an internal Ethics Office. Every piece of content passes through a Compliance Officer, a Lawyer, and Meg—the Ethics Manager. Meg holds a “zero-tolerance policy” capable of blocking workflows entirely if she identifies missing claims, bias, or manipulation.
This governance is rooted heavily in E-E-A-T (Experience, Expertise, Authoritativeness, and Trust). The creator’s “Core Mission” reflects this: “To bridge the gap between technical data and human intuition through collaborative AI… always under human supervision.”
The Homogenization Trap: However, even the most rigorous internal E-E-A-T scoring has a blind spot. If a human operator pushes for pure volume, the system’s output risks becoming a detectable, homogenised pattern that algorithms learn to downrank—because true first-hand experience and authoritativeness often carry idiosyncratic voice and nuance that guardrails tend to smooth out. If the strategy prioritises scale over the end-user, the system will inevitably collapse under Google’s quality filters.
The Honest Review: Is “Agency” Actually Too Powerful?
If we strip away the aesthetics of an AI workforce and look at the system mechanically, the truth is far more grounded: The system is not too powerful; it is just highly leveraged. And leverage cuts both ways. Here is an honest assessment of why replacing a human agency with a synthetic workflow engine still faces hard, unyielding limitations.
1. The Strategy Multiplier Effect (Garbage at Scale)
Agentic tools do not invent fundamentally new business strategies; they perfectly execute the parameters they are given. It is a multiplier.
If the human “CEO” inputs a brilliant, insightful directive, the agents will produce brilliant work at scale. However, if the operator lacks deep market insight, the system will simply automate mediocrity at 100x speed. You cannot automate product-market fit or empathy. The system makes the human operator the single point of failure—if the strategy is flawed, the AI will just distribute that flaw everywhere simultaneously.
2. The Illusion of Genuine Autonomy
Despite the impressive operational layers, these agents operate in predictable, chained functional loops.
The necessity for “Human-in-the-Loop” (HITL) approvals from the CEO is a feature, not a bug. Without a human acting as a harsh, editing gatekeeper, synthetic content workflows inevitably degrade into repetitive ‘AI-speak’. The operator trades the friction of creating content for the friction of constantly reviewing and correcting it.
3. The Relationship and Trust Barrier
Business is fundamentally about human trust. While future iterations of telephony agents may field calls with zero latency, automation solves fulfilment, not connection.
Marketing and client services rely heavily on social capital and empathy. An AI receptionist cannot read the emotional subtext of a frustrated enterprise client or navigate delicate, relationship-driven negotiations over dinner.
4. Economic Asymmetry vs. Platform Fragility
The economics are compelling: AI “employees” don’t require salaries, and compute costs yield massive margin improvements. A skilled individual can punch at the level of a mid-sized team.
Yet, this power relies on a staggering number of external dependencies (Gmail APIs, OpenAI, Anthropic, Twilio, etc.). When these APIs change or are rate-limited, the orchestration breaks down. The operator quickly finds themselves spending less time executing strategy and more time doing software engineering and prompt-chain debugging.
The Final Verdict: A Tool, Not a Leviathan
Codename “Agency” is a monumental achievement in workflow automation. It democratizes execution, levelling the playing field for solo operators. However, it is an incredibly powerful “cyborg” exoskeleton, not a fully autonomous business-in-a-box.
As agentic systems become more accessible, the winners won’t be those with the biggest agent swarms, but those who use them thoughtfully. The ultimate constraint remains exactly where it has always been: the strategic vision, taste, and empathy of the human operator sitting at the terminal.
Although, as I have hinted at previously on Social Media, Agency is a lot more subversive than just a hyper-efficient swarm of AI employees… full details soon, but know this: this system is too big for me to complete the artwork myself. If you have skills, you want them to pay the bills. Look around you and ask yourself why this is so difficult. This is the big idea…

Disclaimer: Agency is a simulation. AI is a simulation! Some people think you are a simulation. Not many people know this aspect of AI. AI cannot do a lot of things people say it can do. For transparency, I need to say it is a simulation. For instance, I have an accountant baked into Agency. I am not an accountant, though, and neither is AI. This data should be reviewed BY YOUR REAL ACCOUNTANT, with the point being you have saved a lot of time collecting, sorting and reviewing the data before the real accountant reviews it. That is the essence of Agency. It is the essence of AI HITL Marketing. AI does the grunt work, the CEO signs the job off, and publishes under their credentials. This article was created using Grok, CLaude and Gemini to give a thoughtful, transparent overview of what Agency, and Agentic systems like it represent over the coming year. For me, at this point, the cyborg apparatus I predicted in my AI SEO ebook last year is now effectively built, and it is just a job of making it slicker and ensuring it is wielded correctly and responsibly by the user.