Insight
OpenAI just bought a consulting firm. Here’s what they can’t buy.
OpenAI can hire people who plug AI into your company. But knowing what a great brand strategy actually looks like in your industry, what works, what fails, and why, only comes from years of doing the work. That experience lives in people’s heads, not in anything OpenAI can write a check for.
Let me back up.
In May, OpenAI launched a new business called the OpenAI Deployment Company and acquired an AI consulting firm, Tomoro, to staff it. The company said the acquisition brought roughly 150 forward-deployed engineers over from day one, and reporting at the time put the venture’s initial investment above four billion dollars. A week before that, Anthropic announced something similar: a partnership with Blackstone, Goldman Sachs, and Hellman & Friedman to launch a firm that embeds its people inside businesses and wires Claude into their operations, starting with companies the investment firms already own.
Read those two announcements next to each other and the future becomes pretty clear. The companies that make the AI are now also the companies that show up at your office and install it. Getting a capable model into your business, connected to your data and running in your workflow, is about to be easy and cheap and available to everyone.
Which sounds like good news, and in one way it is. But if you run a brand, it should also raise a harder question… Once everyone can get AI into their business, and soon everyone will, the model stops being the thing that sets your work apart. So what does?
Plugging it in was never the hard part
Connecting a model to your data was never the hard part. It felt hard for a while because it was new. But it was always going to get easy, and now it has. The hard part is the same thing it has always been, long before any of this: knowing what good looks like.
An AI model can read every brand document your company has ever produced. It can summarize them, compare them, score them against each other in seconds. What it will not do on its own is tell you that your positioning statement is technically fine but would die on a shelf at Target. It will not tell you that your tone of voice document says “playful” while every piece of packaging reads like a pharmaceutical insert. It will not tell you that you have three audience definitions floating around the building and the expensive one is the wrong one.
That kind of judgment, and knowing the right questions to ask in the first place, comes from somewhere specific. It comes from years of building and running a brand, watching some things work and others quietly fail, and slowly developing a sense for the difference. There is no dataset for it. Most of it has never been written down anywhere. It lives in the heads of your best people.
And that is exactly why the labs cannot sell it to you. They can sell you engineers. They cannot sell you twenty years of pattern recognition in your category. Not at scale, not for your specific brand, and not in a form a model can pick up and use out of the box. The engineers are very good at deployment. Deployment is not the same as knowing what to build, or whether the thing you are about to scale is any good in the first place.
To be fair to the labs, they are not pretending otherwise. Their pitch is speed and reliability, getting a capable model into your workflow without the thing stalling out. That is a real problem and they are well positioned to solve it. The mistake would be assuming that solving it settles everything else. It does not. A faster path to generic work is not progress.
Same model, completely different output
Here is the whole argument in one line:
It is not about having the best model. Everyone has a brilliant model. It is about what you feed it.
Point a brilliant model at generic instructions and you get generic brand work. Fast, polished, and on brand for no one. Point that same model at your actual standards, your real sense of who the brand is for and what separates good from good-enough in your category, and it produces something only you could have made. Same model. Completely different output. The difference is entirely the judgment behind it.
This is why the model itself is a trap to fixate on. When everyone has access to the same capable models, the model stops being your advantage. It is table stakes. Better models running on the same generic knowledge all produce the same generic answer, which means the more capable AI gets, the more the difference lives in what you put in front of it. Every improvement the labs ship raises the value of having something worth feeding the machine. That thing is your expertise, made explicit.
So the real divide is between companies that have codified what they know, in a form a model can actually run on, and companies where that knowledge stays locked in a few people’s heads and walks out the door at six every evening. One of those can put its best thinking behind every piece of work, at once. The other is still rationing its best people across the few projects they can personally touch.
What this actually takes
We have been living this at Upland, so this is not a prediction. It is a report from inside the work.
We took our own approach to evaluating brands, the questions we ask, the standards we hold work to, the patterns we have watched break companies, and we have been turning it into a system an AI can run. The honest takeaway: the AI was the easy part. Writing down what we actually know, precisely enough that a machine could apply it without us in the room, was brutal. And it is the same work any brand serious about this will have to do.
It forces you to confront how much of your craft is instinct nobody ever had to explain. A senior strategist looks at a piece of work and knows, in about four seconds, that something is off. Ask them to write down the rule they just applied and you get a long pause. The knowing is real. The articulation is missing, because they never needed it. For twenty years the job was to have the instinct and use it, not to specify it. The machine does not accept instinct. It needs the rule, stated plainly, with the edges defined.
That is uncomfortable. It is also where the value is, and here is why. Every hour spent forcing a vague instinct into an explicit rule makes the instinct itself sharper. You cannot codify a thing you only half understand. The act of writing it down for the machine turns out to be the act of finally understanding it yourself. The brands that skip that discomfort keep their expertise exactly where it has always been: valuable, real, and stuck in people’s heads, where a model can never reach it.
And to be clear, this does not replace your best people. It extends them. It takes a standard that used to live in a few heads and applies it consistently, across far more work than those few people could ever touch by hand. The judgment still has to be right. What changes is how far it reaches. The strategist who used to shape a dozen projects a year can now shape all of them.
What to do about it
So if you are a founder or a CMO watching the AI consulting gold rush, here is the takeaway worth keeping. Anyone can connect a model to your data now. That part is solved, and soon it will be free. The thing that decides whether the output is worth anything is the judgment you put behind it, and that judgment, the most valuable input you have, has probably never been written down. The labs, by making everything else easy, just gave you a very good reason to start.
This is the work we do at Upland. We are foresight-led operators, people who have held the seat, and we help brands turn their own judgment into something an AI can run on, so the technology scales your best thinking instead of averaging it away. We learned how hard it is by doing it to ourselves first, which is also how we know it is worth it.
If you are wondering how to make AI produce your brand’s best work, that is a conversation worth having. It usually starts with a simple question: what do your best people know about this brand that has never made it out of their heads? We are happy to help you find the answer, and to help you write it down.
