Businesses already pay for access to AI agents. The access is usually bundled inside a service, so most people have not noticed the shift happening.
At TJ Digital, our AI workflows are the second most valuable asset we have, behind only our people. We have put thousands of hours into building them, and they now do most of the work across roughly 40 to 50 client campaigns.
Were the 2023 predictions about selling AI access right?
Back in 2023, people were talking about creating an AI version of themselves and charging others for access to it. That sounded crazy to me at the time. Why would anyone pay for an AI trained on one person’s information when they could use ChatGPT, which is trained on basically everything online?
It turns out those people were mostly right. They just had the shape of it wrong.
Nobody is buying an AI clone of me. Businesses are paying for systems built to do specific, valuable work, with the AI doing most of the work.
What are you actually paying for when you hire an AI-powered agency?
The number one asset at our agency is the people. The number two asset is our Claude skills, a library of markdown documents that run our workflows. I never would have guessed a set of text files could be worth this much, but thousands of hours of building and refining went into them.
Every service we provide still involves a human. Most of the actual work is done by Claude, ChatGPT, or Gemini. Over time, more of that work shifts to these agentic workflows, and our people spend more of their time on what they are best at, which is building relationships and applying judgment where it matters.
That blend of people and AI is how we deliver roughly four times the workload at the same rates.
So when you hire us, part of what you are paying for is access to an AI agent. We are open about that. We decided early on who builds the AI systems and who runs them, and we only put AI on a task when it does a better job than a person would.
That standard is also why using AI for content works for us when it backfires for most businesses.
Why are clients asking to license AI workflows directly?
More and more people are asking if they can skip the service and just get the AI. They will say something like, “How about you train my team to use your workflows?”
This tracks with a broader move. Companies want AI inside their own operations, and even the biggest AI labs are now embedding AI directly into businesses. The instinct is the same whether it comes from a Fortune 500 or a small business owner watching one of my videos.
I don’t have a clean way to do that yet. And honestly, even if I did, I have no idea how I would price it.
How do you price and sell an AI agent?
This is one of those questions the whole economy is going to have to work out. I don’t have an answer, so I will lay out both sides the way I see them.
On one side, the thing being sold is a series of markdown files, maybe with some database connections or MCP servers behind them. Most of the value sits in those text files, which can be copied and shared in seconds. A lot of people are going to be hesitant to attach a heavy price tag to that.
On the other side, these AI agents now produce very high-quality work. A workflow only goes into production once it outperforms a person, so by then it is doing something a skilled human would charge real money for. That is worth paying for, regardless of how simple the underlying files look.
| Why an AI agent can look cheap | Why an AI agent is worth paying for |
| It is mostly a set of markdown files | The output rivals or beats skilled human work |
| Those files can be copied or shared in seconds | Thousands of hours went into building and testing them |
| Anyone can open ChatGPT and start typing | The workflows encode hard-won judgment about what works |
| The underlying model is available to everyone | The system decides what to do and checks its own work |
The pricing models are starting to shift to match this. Charging by the hour stops making sense when an agent finishes a task in seconds, so firms are testing a few approaches:
- Subscriptions and licensing, where you pay for ongoing access to the workflow the way you pay for software
- Outcome-based fees tied to the value created, like the hours saved or the revenue generated
- Workflow as a service, where the prompts stay private and you get the result without the method
There is a catch on the cost side too. AI is not automatically cheaper. A premium model running a complex task can cost more than the human labor it replaces, which means nobody is going to pay extra for AI work unless the output is clearly better or the savings are real.
For now, I don’t have answers. This is something we are all going to have to figure out, and I think it is one of the more interesting open questions in business right now.
What does this mean for business owners right now?
The practical move is to pay attention to who is doing your work and how. If an agency is using AI in every workflow, that should show up in your pricing and your output, and it should not stay hidden so someone can keep the savings to themselves.
Can you buy access to an AI agent today?
Yes, though almost always as part of a service. The businesses furthest ahead deliver the AI’s work to you while a human steers it and handles the judgment calls.
Why pay for a custom AI agent when ChatGPT is free?
The value comes from the workflows and judgment built around the model. The model on its own is the easy part. ChatGPT will write you a blog post, but the system that decides what to write, checks it, and makes it perform is where the thousands of hours go.
How much does it cost to license an AI workflow?
There is no standard price yet. The market is moving away from hourly billing toward subscriptions and outcome-based pricing, and where things land will depend on how much measurable value the workflow creates.
We run AI search campaigns for roughly 40 to 50 businesses, with AI doing the work and a human steering every account. Reach out to TJ Digital to see what that could do for your business.