Breakfast of Consultants

Listen…

There is a company in California called OpenAI. They’ve spent billions of dollars and enough electricity to power a medium-sized European country to build a brain in a box. They told us this brain was going to change everything. It was going to write our code, cure our diseases, and probably walk our dogs.

Instead, they just hired McKinsey.

OpenAI recently announced the “Frontier Alliance,” a multi-year partnership with Boston Consulting Group, McKinsey, Accenture, and Capgemini.

The goal? To help enterprises figure out how to use their no-code AI agent platform.

Let that sink in for a moment. The most hyped, purportedly paradigm-shifting technology in the history of human commerce cannot sell itself. It cannot explain its own value. It needs a 28 year old consultant, armed with a two-by-two matrix and a Patagonia fleece, to explain to Fortune 500 CEOs why they need it.

If you work in product, this news should make you stop dead in your tracks because what OpenAI is quietly admitting is something we’ve all been whispering loudly for the last eighteen months: AI is missing the mark.

Massively.

The Pilot Purgatory

Here is how enterprise AI works right now:

A CEO reads a magazine on an airplane. The magazine says AI is the future. The CEO lands, calls their VP of Product, and says, “Put the AI in the thing.” The product team panics. They build a chatbot. They stick it in the corner of the app. It costs a fortune to run. It occasionally hallucinates and offers a customer a refund for a product they never bought. The company issues a press release. The stock bumps up 2%.

Then, nothing happens.

The pilot fails to scale. The customers don’t use it, because it turns out people don’t want to have a deep, generative conversation with their accounting software. They just want to balance their books and go home to their children.

This is why OpenAI had to call the consultants. Because fragmented tooling, siloed data, and bespoke integrations are ruining the fantasy. You can’t just sprinkle an LLM on a broken business process and expect magic. A bad process automated by a supercomputer is just a bad process that happens at the speed of light.

A Very Expensive Hammer

Let’s get this straight: AI is not a strategy. It is a commodity. Like electricity, or gravel, or anxiety.

We have spent the last two years pretending that if we just buy enough compute, the strategy will magically reveal itself. That the machine will somehow figure out our business model, fix our supply chain, and write our Q3 OKRs.

But a tool is only as good as its application. If you give a nail gun to a golden retriever, you don’t get a house. You do get a lot of noise and a trip to the vet.

This is the reality of the enterprise right now. We have handed the most powerful cognitive nail gun in human history to middle managers who don’t even know where the studs are, but have been handed a mandate that is akin to “Do maths." Simply put, they lack a fundamental strategy for why they are doing any of this.

Application requires intent. It requires looking at the unglamorous, fleshy parts of your business and deciding exactly where a probabilistic word-calculator can actually reduce friction. That takes time. It takes discipline. And it takes strategy…a word we seem to have largely replaced with “prompt engineering.”

The Illusion of the Automated Workforce

Let’s rewind the tape. When the consumer-focused AI applications started to stall, when regular people realized they didn’t actually want to pay twenty dollars a month to have a chatbot write a mediocre email or plan a generic vacation, the narrative had to pivot. If the tech companies couldn’t sell to the masses, they would sell to the bosses.

Suddenly, we were promised “agents.” We were promised autonomous, tireless digital workers that would seamlessly integrate with our systems, eliminate the need for bloated payrolls, and reduce the human workforce to dust. We were told the trajectory of AI was a steep, unstoppable curve toward artificial general intelligence. But the OpenAI and Big Consulting partnership reveals a much more boring reality: the trajectory of AI is hitting the hard wall of human friction.

Because here is another dirty little secret: not a lot has changed. We still need competent humans.

Spoiler.

The plumbing of the global economy is still held together by Excel spreadsheets from 2014 and a guy named Gary who knows the legacy database passwords. The media loves to run breathless headlines about AI replacing millions of jobs, but there is exactly zero evidence that AI can replace a job without context.

An LLM can write a beautifully structured Python script or generate a sparkling Q3 marketing plan. But it doesn’t know why the Python script needs to bypass a broken API that nobody has bothered to fix since the Obama administration. It doesn’t know which VP will throw a fit if the logo is on the left instead of the right.

Context is everything, and right now, context is a closely guarded hostage.

The Context Hoarders in Layoff Land

In our current hiring climate, where tech workers are exhausted, companies are running endless “efficiency exercises,” and everyone is quietly terrified of losing their health insurance, institutional knowledge is the ultimate moat. Employees know that their undocumented, duct-taped workflows are the only thing keeping them indispensable. So, they hoard the context. They guard their messy, idiosyncratic business processes like dragons guarding gold.

This is the brick wall that OpenAI and their new consulting buddies are going to hit at a hundred miles per hour. No consulting firm, no matter how many billable hours they log, can wrangle that information free. You can’t send a junior analyst to interrogate a veteran supply-chain manager and expect them to happily hand over the secret sauce so a chatbot can take their job. It defies human survival instincts.

Without that hoarded context, the smartest AI in the world is just a very fast, very confident idiot. A bad process automated by a supercomputer is just a bad process that happens at the speed of light.

Why You Still Need the Damn Tractor

Look. Just because the tech billionaires lied to you about the impending singularity does not mean you get to ignore the technology entirely. Should we throw the servers into the ocean and go back to carbon paper?

No. Don’t be ridiculous.

AI is not your new Chief Marketing Officer, and it is not a magical replacement for Gary in IT, but it is a phenomenally powerful tractor.

When John Deere started mass-producing tractors, it didn’t eliminate the need for farmers. It eliminated the need for farmers to break their spines plowing dirt by hand. It let them farm more, and it let them farm better.

Your company is full of digital dirt. You have highly paid professionals spending half their waking lives doing robotic, soul-sucking tasks: formatting spreadsheets, cross-referencing PDFs, writing repetitive status updates, and summarizing meetings that should have been an email in the first place. This is a profound waste of human life.

You must still use AI to streamline your processes, because a human brain is too precious to be used as a copy-paste machine. Use the LLM to shovel the dirt. Give your employees the tractor. If you deploy AI as a tool to remove friction rather than a scheme to remove payroll, you will actually see the productivity gains everyone else is merely pretending to have.

The Boring Business of Babysitting the Brain

Here is where the actual, tangible value of AI sits right now: It isn’t in the models…It is in the leash.

The greatest value in AI today is running governance and building structured strategies for deployment. Everyone is obsessed with what the brain can do, but enterprise value is entirely dependent on ensuring what the brain won’t do. It won’t leak customer data. It won’t hallucinate a legally binding contract. It won’t accidentally email the payroll spreadsheet to the entire marketing department.

This is precisely why OpenAI had to call the consultants. They realized that you can’t just drop an omnivorous, unpredictable intelligence into a rigid corporate structure and expect it to magically align with compliance standards. You have to build fences. You need a strategy. You need governance frameworks, risk assessments, and deployment structures that map to the agonizingly slow reality of human “bureaucracy.”

This is where I’d like to make a pithy statement about layoffs, but I will hold my tongue.

The companies that win the next five years wont be the ones with the smartest, most “agentic” AI. They will be the ones that master the deeply unsexy art of AI governance. They will figure out how to safely pipe their data into the machine without violating privacy laws.

What This Means for Product

If you are a product manager, a designer, or an engineer, the Frontier Alliance is a massive, blinking warning light for how you build.

• The tech is no longer the product: OpenAI is openly admitting that model intelligence is not the bottleneck anymore. Organizational design is the bottleneck. If the smartest AI in the world needs corporate change management to be useful, your little wrapper app certainly isn’t going to survive on its own. You have to build for the messy human reality, not the technological capability.
• Stop building features looking for problems: We have spent two years shoving generative AI into every crevice of software because we could, not because we should. If you don’t deeply understand the user’s workflow, and the hoarded context behind it, no amount of AI will save your feature from abandonment.
• The Emperor’s New Tech Stack: The trajectory of AI isn’t democratizing intelligence; it’s just centralizing IT budgets into the hands of the people who already run the world. OpenAI’s engineers are going to sit alongside BCG and charge millions to tell your boss how to run your tech stack.
• The Product is now the brand: When the underlying technology is a commoditized brain rented by the millisecond from a server farm, you no longer have a technical moat. Everyone has the same magic trick. What you have is trust. Your brand isn't just your logo or your clever marketing copy anymore. Your brand is the actual, tactile experience of using your product. If your software feels like a chore, no amount of AI pixie dust will save you. You must build something people actually like.

The machines are very smart, but the humans are very tired.

We were promised a revolution. Instead, it is the same circus, just with slightly smarter elephants. If you are building products today, ignore the noise. Ignore the alliances. Stop worrying about whether your AI agents are “agentic” enough.

Look at the person using your product. Are you saving them time? Are you navigating their lack of context? Are you helping them get away from the screen faster so they can go look at a tree or kiss their spouse?

If not, all the consultants in the world can’t save you. I promise I didn’t make this up, and here are some sources for you to check if you’d like.

Sources

TechCrunch (February 23, 2026): https://techcrunch.com/2026/02/23/openai-calls-in-the-consultants-for-its-enterprise-push/

OpenAI Announcement (February 23, 2026): https://openai.com/index/frontier-alliance-partners/