With new AI models multiplying faster than browser tabs, friends, colleagues, and clients are asking us how we’re using it in our work. So, here we “think out loud” and share our observations and experiences. Please weigh in with yours!
The danger is not bad AI. The danger is believable AI.
A recent social post sharpened our thinking. It cited Gartner research to suggest that consumers were consulting AI but not using it to make purchase decisions.
That was not what the research said. People were using AI to inform decisions. They simply were not completing the transaction through an AI engine or agent.
The post sounded polished enough that the distinction was easy to miss. That is the risk. If you are not already close to the subject, AI can look brilliant even when it is wrong.
AI has lowered the cost of sounding knowledgeable. Anyone can produce a credible summary, framework, or point of view in minutes. But speed is not mastery, and a strong dossier is not proof of expertise.
AI is minting experts faster than it is creating them. Eventually, someone who knows the subject walks into the room. Once people realize you did not understand what you repeated, they begin to question everything else you say.
AI will loan you expertise. Reality will collect the debt.
Systems create time. People create meaning.
We see the greatest value in AI when it is designed into systems.
It can organize research, transcribe interviews, compare sources, assemble drafts, flag claims, complete tasks through agents, and adapt work across channels.
The point is not more output. It is more time for investigation, creativity, planning, and collaboration. Technology should extend human thinking without diluting it.
Systems create time. What we do with that time is create meaning.
Control the machine.
Our clearest internal directive is simple: Control the machine. Do not let it control you.
We define the objective before asking AI how to achieve it. We provide context, sources, standards, and audience perspective. Then we review the result as critically as we would review a colleague’s work. We remain responsible for every conclusion, recommendation, and piece of content carrying our name.
The machine does not know what matters to the client, what could damage trust, which assumption is dangerous, or what the person who is not in the room might hear.
The machine does not harness itself. That is our job.
Every prompt reveals a way of thinking.
The moment you open an AI tool, you begin taking it down a path.
Your question reflects what you know, what you want, and what you notice. Each correction, example, disagreement, and instruction makes the exchange more specific to your worldview.
AI may offer five possibilities, none quite right. You may combine parts of two and discover a sixth idea that did not exist before the exchange.
That is where the value begins.
If you take the first answer, AI is a vending machine. If you shape the conversation, it becomes a collaborator.
Systems do not replace people. They reveal where people matter.
We automate research, comparison, transcription, organization, production support, and distribution. But every process has moments when a person must enter, review, challenge, or decide.
A good AI-enabled system makes those moments clearer.
Our newsletter, The Architecture of Meaning, is one example. Every issue begins with a conversation about what we are seeing, what changed our minds, what clients are asking, and what remains unresolved. Nothing is drafted until that conversation is complete.
AI organizes the ideas, searches source material, and proposes a structure. A human redirects or approves it before drafting. We then edit, expand, correct interpretation, and remove anything that merely sounds good.
AI makes the process more efficient. Human interaction makes the work worth reading.
The work that requires us.
Nobody needs a person to hand over unedited AI output. Anyone can generate that for themselves.
The human contribution begins with the response:
Does this say anything?
Is it true?
What is missing?
Which assumption should be challenged?
What would a customer, employee, investor, or critic hear that the writer missed?
Sometimes the answer is, “This is exactly right.” Sometimes it is, “This sounds impressive but says nothing.” Sometimes it is, “Absolutely not. You would light the whole city on fire with that recommendation.”
And sometimes useless output sparks a better idea.
Distinguishing meaning from noise, representing people who are not in the room, and turning possibility into direction is where we matter.
Our value is not just the output. It is what shapes it.
AI can help produce words, research, frameworks, and recommendations.
Our value comes from what shapes them: decades of experience across startups, enterprises, institutions, public-sector work, leadership transitions, market shifts, successes, and failures. Experience changes what we notice, the questions we ask, the risks we recognize, and the patterns we see before they become obvious.
This is the work behind the words.
The same is true of AI. The output is visible. The value lies in what shaped it.
Our expertise cannot be downloaded.
The future changes. Our method does not.
Tools, models, prices, and platforms will change. Some will disappear entirely.
We cannot build an advantage on access to a particular model. We can build one on the way we think.
AI can support strategy, positioning, messaging, expression, relationships, meaning, and momentum. It does not eliminate the need for any of them.
The tools will evolve. Our responsibility to think clearly, judge carefully, and remain accountable will not. No one is operating from a finished playbook. We will keep testing, refining, and questioning how we work.
Where does human judgment create the greatest value in your work?
What have you automated that improved the outcome?
How do you test whether AI-generated work is true, meaningful, and worth using?
We are not just building an AI-enabled company. We are building a way of thinking about one.
Read more about The Outfront Method.
