Every Client Is Asking About AI. Here Is What We Tell Them.

In the last six weeks, four of our clients have asked us some version of the same question: "What should our AI strategy be?" One board apparently put it on the agenda by name. This is what happens when a technology gets a hundred million users faster than anything in history. Nobody wants to be the executive who missed it.

We take the question seriously. We just refuse to answer it as asked, because "what is our AI strategy" is a category error. You do not have a spreadsheet strategy or an email strategy. You have a business strategy, and tools either serve it or they do not.

The reframe we use

When the question comes up, we walk clients through three narrower questions that actually have answers.

Where does language do work in your business? ChatGPT and its successors are, today, machines for producing plausible text quickly. Map every place your company produces text at volume: support macros, product descriptions, internal documentation, ad variants. That map is your realistic opportunity surface. For Northwind, a retailer with forty thousand SKUs, product copy is a genuine bottleneck and a genuine candidate. For a brand whose voice is the product, the calculus is different.

What is your tolerance for being wrong in public? These models are confidently inaccurate. That is manageable when a human reviews every output and catastrophic when one does not. Any use case should be sorted by blast radius before it is sorted by savings.

What data are you comfortable sending to a third party? Legal has opinions. Ask them early, not after a well-meaning team has pasted the customer database into a chat window.

What we tell them not to do

• Do not appoint a Head of AI in January 2023. You will hire for a job description that will be wrong by June.
• Do not ship an AI feature because your competitor shipped one. Their press release is not evidence that it works.
• Do not confuse a demo with an operation. Everything demos well right now. Very little is deployed with monitoring, review, and a rollback plan.

The honest answer about creative work

Clients also ask, sometimes gently, whether this changes what they should pay agencies. Fair question. Here is our honest answer: the floor is dropping and the ceiling is not moving. Acceptable copy, competent layouts, serviceable strategy decks: all of that is getting cheaper, fast, and agencies whose business was volume production of acceptable work should be nervous. We are not being brave when we say this. We are being descriptive.

But the value of acceptable work was already collapsing before the machines showed up. What clients actually pay a studio for is judgment: which of the two hundred possible directions is right, what to kill, what a brand should refuse to say. Nothing we have seen from any model, and we have pushed on them hard, exhibits taste or accountability. When the output is wrong, no model takes the call.

Our working position, dated for the record

It is January 2023 and everything below is subject to revision, which is rather the point:

• We use these tools daily for drafts, summaries, and thinking out loud. We disclose that when asked.
• Nothing generated ships to a client without a human who is accountable for it.
• We advise clients to run small, reversible pilots with clear review gates, and to write down what they learn.

The companies that will look smart in two years are not the ones with the boldest AI announcement this quarter. They are the ones quietly building the muscle: where it helps, where it fails, and how to tell the difference. Strategy is still what it always was, deciding what not to do. The tools just got louder.