Could ChatGPT, Claude or Gemini replace your marketing agency?
For a lot of the work you are paying for, I think the answer is yes. Someone still needs to manage it, know what good looks like and make sure it works for your business. That could be you, someone on your team or an agency that knows how to use these tools properly.
But here is the question I would ask: if AI now does much of the work, what happens to the bill?
Are you getting more value from the power of AI? Or is your agency getting all of it?
I run an agency myself. These are questions we should be willing to answer, too.
When AI cuts agency production time, what happens to the fee?
We used to charge for creating an article because there was work involved. Someone had to write it, revise it, find a suitable photo, look through the website for relevant links, add those links and post it. A brochure meant more writing, design and preparation for print or the web.
Our fees reflected the time those jobs took.
Now think about a job that used to take five hours. Suppose a well-set-up AI process brings it down to ten or fifteen minutes, with the quality you need. That is an example to consider, not a promise about every job. Research, revisions and approvals can still take time.
Does the agency come back and say, “We can do more for you now”? Does it lower the price? Does it use the time to improve your marketing?
Or do you receive the same thing, at the same price, while the agency quietly keeps the difference?
It feels a little like The Wizard of Oz. Once you have seen behind the curtain, the conversation changes. You want to understand what is actually being done for you.
AI can replace more marketing production than people realise
ChatGPT Work can research topics, use connected apps and create finished documents. Claude can create and edit documents, presentations and spreadsheets. Gemini Canvas can help create documents, slides, apps and web pages. Features and access vary, but the production tools are already here.
With approved facts, your brand voice and suitable access, an AI process can help research an article, draft it, prepare images, suggest internal links and assemble a page for review.
Someone still has to check the claims, image rights, layout and final result. A brochure that looks good on a screen may still need print preparation. Connecting a publishing tool also takes setup and permission. Those are real tasks. They should be explained, rather than hidden inside a vague monthly fee.
There is evidence behind the productivity claim. A 2023 experiment with 453 professionals found that ChatGPT reduced average time on selected writing tasks by 40% and increased assessed quality by 18%. That study did not test an entire marketing department or prove that every agency job is faster. It does show why clients are questioning the old time estimates. Read MIT’s account of the study.
The client should share in the benefit
This question is already being asked by companies with much bigger budgets than a local business.
In a 2024 WFA and MediaSense survey of more than 80 multinational advertisers, 58% expected to pay less where AI was used. Yet 61% expected agency fees overall to rise, particularly for strategic and technical talent. These are large advertisers, not a survey of small Canadian businesses. Still, the distinction makes sense: routine production gets cheaper while valuable expertise can earn more. WFA research.
An agency is allowed to make a profit. Building a reliable process takes investment, and knowing how to solve a problem quickly has value. A fixed price also pays for an agreed result, not necessarily a set number of hours.
But if hours were the reason for the price, and those hours have largely disappeared, I think the client deserves a fresh conversation.
That could mean a lower fee. It could mean more useful work, faster delivery, better testing or a system your team can keep using. The benefit should show up somewhere you can see.
Would you rather have ten articles or more customers?
Could an agency offer ten articles for what one used to cost? Possibly. But ten articles nobody needs will not help you much.
I would rather see the agency use that capacity to answer the questions your customers actually ask, improve important service pages and find out which messages produce inquiries. I raised a similar concern in Is Your Social Media Stuck in Groundhog Day?. More of the same can just give you more of the same results.
The agency should explain what it chose to improve and how you will judge whether it helped.
Who is actually doing your agency work?
The owner may talk about AI. The salesperson may give a great presentation. But does that knowledge reach the people working on your account? Can the team do what the pitch promised?
Junior staff can do excellent work with the right skills and support. A senior title alone proves very little. I would want to meet the people doing the work and understand who guides them, checks it and takes responsibility.
I learned something about this while working for Arlene Dickinson at Venture Communications. I remember another major client congratulating Arlene and the agency on a large new client. They also made sure their own team would not change. They wanted to protect the experience and continuity they were paying for.
That was smart. They were asking what the agency's growth would mean for their account. It is a lesson I would carry into an AI conversation today: whose ability are you buying, and will those people still be there after you sign?
Could you manage the AI yourself?
For many owners and founders, I think you can bring meaningful work in-house. Try a recurring task you understand. Give AI the brief and approved information, review the result, and count your time, corrections and tool costs. Compare that with what you currently receive.
Coaching or a setup project may be enough. If you need help connecting inquiries, customer records and follow-ups, our workflow automation service addresses that work. A provider should explain what it builds, checks and maintains.
Ask your agency these questions
You do not need to understand every model or plugin. Bring these questions to your next meeting:
- Which parts of our work now use AI? Ask them to walk through a recent deliverable, including the checks and revisions.
- What has changed in the time and cost? Separate production, setup, management, software and advertising spend.
- How are we benefiting? Ask for a concrete change in price, useful output, turnaround or results.
- Who will actually do our work? Ask for names, roles, relevant skills and examples. Meet the delivery team, not only the people presenting.
- Does AI knowledge run through the business? Ask about leadership, sales and delivery. Is there time for training and practice, or does it depend on one person?
- What happens when our team changes? Agree how substitutions will be discussed, who maintains standards and how knowledge of your business is handed over.
- What do we own if we leave? Clarify access to accounts, content, data and the systems you paid to build.
Agree on a starting point and review the result together. Qualified inquiries, bookings, sales and time saved tell you more than a report showing how many items were produced.
The agency has to earn its place
We have gone AI-first in my own agency. I research, practise and share what I learn. That has meant changing how we work and who is suited to doing it. A chatbot subscription alone would not have been enough.
I think much of the traditional production agency can be replaced by AI tools managed well. A good agency can still earn its place by helping you achieve something you could not easily achieve alone.
For a startup, that may mean operating leaner. For a local business, it may mean more calls and bookings. For an association or larger company, it may mean a team that gets more useful work done.
At Salient AI, our AI consulting starts with what you want to accomplish. Apply the same test to us.
At the end of the day, are we using AI to move your business forward, or just to make our own work more profitable?