AI is everywhere now, and we know the thought has probably crossed your mind: if AI can write copy, put together a strategy in seconds, generate a design (some more convincingly than others) and code a website, what exactly are agencies here for?
AI has reached the point where not using it at all can mean falling behind. In McKinsey’s global survey on AI adoption, 71% of respondents said that their organisations regularly used generative AI in at least one business function, with marketing and sales among the departments where it emerged most often (Tinkoff et al., 2025).
So yes, we use it too, and we’re glad it’s here.
It can help us interrogate a thought or sort through information that we would otherwise have spent hours on, doing ourselves. In the right hands, there is no denying that it can be genuinely useful.
So, to be clear, we’re not here to argue that agencies should pretend AI doesn’t exist or cut it from workflow completely. The evidence gives us very little reason to. And frankly, that would be a fairly strange hill for an agency in 2026 to die on.
The more interesting question is what happens once everyone has access to the same technology.
Having AI isn’t the same as knowing what to do with it
More than 60% of US agency decision-makers surveyed by Forrester said their agencies were already using generative AI, while another 31% were dabbling in the idea thereof. Creative ideation was one of its most common uses (Forrester, 2024).
In other words, access itself is becoming rather unremarkable. Think back to the first time you typed a prompt into an AI tool and watched it produce something in seconds. It felt like magic. Now? Most of us barely blink. Anyone with an internet connection can open a tab and relish in a buffet of AI tools, each designed for a different kind of task.
And that changes the agency conversation.
The value no longer lies in the tool itself. More increasingly, the value lies in discerning where the tool belongs in the creative process, what requires human scrutiny, and who takes responsibility for what eventually reaches a client.
When everything starts to sound a little too familiar
There is another tension worth paying attention to.
As more people lean on the same technology, we’re starting to notice how much everything can sound alike. The same phrases resurface.
It almost makes you think of those cartoons you watched as a kid, where a character would throw on a disguise, maybe a pair of glasses or a wildly unconvincing moustache, and somehow fool everyone except the person watching.
That’s where the trade-off comes in. Yes, AI can help us produce better-quality work, but it can also start nudging us towards the same kinds of ideas and language. For an industry where originality and freshness are part of the job, that creates a real tension.
As AI fatigue slowly starts to creep in, people are more and more starting to gravitate towards the things that connect them back to what feels lived.
Like those moments when a certain phrase takes you straight back home to a lost memory, or a design reminds you of someone you haven’t thought about in years.
AI can recognise patterns behind those feelings. It has never had the experience itself.
We have. We’re people too.
That’s where Authentic Intelligence comes in
All of this has us thinking about another kind of AI: Authentic Intelligence.
For us, Authentic Intelligence means using AI without handing our thinking over to it.
It means accepting that the technology can make people work faster, but also recognising that human judgement is more important than ever.
For a content strategist, human judgement lives in the thinking, the language and the story behind it. A designer can’t stop at “it looks good”. A web developer can’t stop at “it works”. Human judgement is what identifies the gap between technically acceptable work and something that makes the audience feel understood. And that feeling of being understood is, ultimately, at the heart of being human.
Accountability still belongs somewhere
There isn’t one set-in-stone rulebook for how creative agencies should approach generative AI.
There are, however, useful frameworks.
The AI Risk Management Framework, for example, from the US National Institute of Standards and Technology (NIST), is built around four functions: Govern, Map, Measure, and Manage. Its generative AI profile uses these to address risks like privacy, intellectual property, information integrity, and human oversight (National Institute of Standards and Technology, 2023).
So accountability cannot be outsourced to an AI model via a prompt. Someone should still take ownership of the output.
Which led us to a simple question: what would the NIST principles look like inside the day-to-day reality of a creative agency?
The CMXM Authentic Intelligence Framework
Adapted from NIST’s Govern, Map, Measure, and Manage functions (National Institute of Standards and Technology, 2023).
The principle behind it is simple: AI should empower teams to build on their existing strengths without compromising the quality of the work or quietly adding to people’s workload.
| NIST function | Agency application | Protect the people | Protect the work |
|---|---|---|---|
| GOVERN | Set the rules | Be clear on why AI is being used. Make sure teams know how to use it responsibly. | Keep AI use within approved tools and clear data boundaries. Keep a human accountable for the final output. |
| MAP | Know where AI belongs | Use AI where it makes sense for the task, not just because it’s available. | Decide how much human oversight a task needs before AI is used. |
| MEASURE | Check how it’s working | Check whether people can recognise when AI is wrong or when it shouldn’t be used. | Test AI-assisted work against the agency’s standards for quality and accuracy. |
| MANAGE | Act on what you learn | Change the process or provide more support when the current approach isn’t working for the people using it. | Increase human review, or stop using AI for a task, when the work no longer meets the required standard. |
A practical AI risk management framework for creative agencies, adapted from NIST.
None of this is particularly extraordinary or futuristic. That might be the point.
The responsible use of AI is less spectacular than people imagine, and more like a series of sensible, and rather ordinary, questions: Who checked this? Where did that information come from? Would we put our name on it?
The questions are human because the responsibility should remain human.
A point of view is still the one thing AI can’t generate for you
AI will keep improving. By next year, it will probably be doing things that we consider incomprehensible today. And, somewhat ironically, there’ll probably be another new tool at your disposal by the time you finish reading this.
Human creativity isn’t becoming obsolete; it’s making the way we use technology more important than ever.
For us, Authentic Intelligence means taking what AI is good at and holding onto the things that make us who we are: wisdom and judgement gained through experience, cultural understanding, curiosity, memory, and the peculiar associations a human mind makes when one unrelated thing reminds it of another.
Isn’t it cool to be human? To carry around your own little database and point of view from a life that belongs uniquely to you?
So don’t forget that the next time you open an AI tab. Bring your own mind with you. There’s more meaning in it than you may ever realise.
Still wondering what agencies are here for? We have a page for that.
Explore our services ➔
Sources
Forrester (2024). Forrester: US agencies are currently leading generative AI adoption. [online] Forrester. Available at: https://www.forrester.com/press-newsroom/forrester-us-agencies-are-currently-leading-generative-ai-adoption [Accessed 17 Sept. 2026].
National Institute of Standards and Technology (2023). Artificial Intelligence Risk Management Framework. Artificial Intelligence Risk Management Framework (AI RMF 1.0), [online] 1(1), pp.20–33. doi:10.6028/nist.ai.100-1.
Tinkoff, D., Van Der Veken, L., Chui, M. and Balakrishnan, T. (2025). The state of AI: How organizations are rewiring to capture value. [online] McKinsey. Available at: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai#/ [Accessed 17 Sept. 2026].
