The Two Noons

On a Sunday in November 1883, clocks across North America struck noon twice.

It wasn't a fault. It was a decision made by the railway companies. Not the government. Running rival trains across a continent on dozens of incompatible local clocks was causing collisions and killing people, so the railroads carved North America into time zones and reset every station clock at once. They called it the Day of Two Noons. Washington didn't make standard time the law of the land for another 35 years. Fierce competitors agreed the rules of the road first, because the cost of not agreeing had grown larger than the cost of cooperating with the people they were trying to beat.

A little over a decade ago, the same thing happened to the plug on your desk. In 1995, seven rivals (Compaq, IBM, Intel, Microsoft, DEC, NEC and Nortel) sat down and agreed a single connector to replace the tangle of serial, parallel and PS/2 ports that made every device a guessing game. They called it Universal (the U in USB) even though, it has to be said, they and others spent the next thirty years quietly re-fragmenting it into a mess of cables that look identical and do different things. But the handshake held. You can plug almost anything into almost anything. That boring agreement unlocked a trillion dollars of devices built on top of it.

This is the pattern every great technology wave follows. First a period of "let a hundred flowers bloom" proliferation. Then a moment when even bitter rivals agree a standard, because you cannot build anything tall on ground that keeps moving. Side of the road to drive on. Voltage. Frequency. The internet's common handshake. The standard is always the least glamorous part of the story. But it is also the part that decides who gets to build, and on whose land.

We are living through exactly that moment in marketing right now. Developing the USB of an agentic marketing world. But people aren't talking about it enough because it doesn't make pictures.

Here is a way to date our recent past. Amongst the gold, silver and bronze being handed out, Cannes Lions unintentionally crowns one word each year that has dramatically higher frequency than the rest, up and down the Croisette. Programmatic won a little over a decade ago. Generative AI won Cannes 2023, and held the crown through two festivals of synthetic films and infinite variations. This year (the festival opens in under three weeks) the word will be agentic. And underneath the agentic conversation, almost unremarked, the actual plumbing got agreed.

That plumbing has an ugly name: MCP, the Model Context Protocol. You will need to learn it, the same way you eventually learned programmatic, so let me make it plain.

For all of AI's apparent magic, getting a model to do something in another system, such as run your Google Ads account, check your inventory or pull your analytics, used to require someone to hand-build a custom connection for every single pairing. Every model to every tool, one bespoke bridge at a time. It was the serial-and-parallel-port mess all over again: brittle, expensive, and impossible to scale.

MCP is the agreement that ends that. It is a single, common standard for how an AI agent talks to any tool or any data source. Build the socket once, and any compliant agent can plug in. Its own designers reached for exactly the comparison I've been making: they describe MCP as a USB-C port for AI.

So when you read that Google, Meta, Amazon and TikTok have each "shipped an MCP server," translate it like this: all four of the dominant ad platforms have now fitted a standard socket to their advertising systems, so that an AI agent (powered by Claude, ChatGPT, whatever you like) can plug straight in and operate the account. Set the budgets. Change the bids. Build the campaign. Shift the targeting. Without a human clicking through the dashboard at all.

The story this week is not any one of those launches. It's that the set is now complete. When one platform opens a socket, it's a feature. When all four agree the same socket, it's a standard and a standard is a different kind of event entirely. That is the Day of Two Noons for media buying.

And here's why it matters more than three years of generative dazzle. We spent those years mesmerised by AI making things like copy, images, a thousand versions of the same ad. Useful, but largely a distraction from where this technology is genuinely strong. What AI is actually, reliably good at is pattern recognition, matching, and executing against those patterns at machine speed which is precisely what media optimisation is. MCP moves the spotlight off making and back onto doing. The shine moves from the shop window back to the plumbing (and this should give you a better understanding of the LiveRamp acquisition by Publicis). That's the whole story.

Before anyone clears the trading desk, the honest counter-case is this. And I get to make it as someone who has built in this gap for a living.

A protocol shipping is not the same thing as an autonomous buyer you would trust. The distance between "Meta fitted a socket" and "I let an agent run my budget unsupervised at two in the morning" is the entire change-management problem, and it is enormous. Something close to 60% of AI pilots fail, and the ones that work take six to nine months of guardrails, supervision and slow trust-building before anyone hands over the keys. The provocateurs selling "the buyer role is gone in eighteen months" are describing a press release, not a P&L.

We've seen this movie at the level of language, too. Slapping "agentic" on things that already worked fine. A socket is real and important. A socket is not a chauffeur. Hold both thoughts.

MCP is only the first standard to fall into place, the backbone if you like, the layer that lets an agent reach a tool at all. Tellingly, it went the open way: Anthropic built it and handed governance to the Linux Foundation, owned by no one, much as TCP/IP was.

The layer above it, where an agent actually pays, completes the transaction and settles up with another agent has no agreed standard at all. That part is still firmly pre-1883. OpenAI and Stripe are pushing one protocol. Google and sixty-odd partners are pushing another. Coinbase, the card networks and others are pushing more. Merchants are already being told they'll need to support two or three at once, because there's no shared clock yet and the prize, by McKinsey's reckoning, is a three-to-five-trillion-dollar agentic economy by 2030. Nobody cedes a standard that size without a fight.

So MCP is the media backbone getting its Day of Two Noons. The money is still a brawl. More handshakes are coming and the industry is, visibly, growing up in public.

When the railroads reset their clocks in 1883, the second noon they created belonged to everyone. It became shared infrastructure or a clock the whole country could set its watch by, regardless of who laid the track. That's one kind of standard.

History gives us two. There's the liberating kind like TCP/IP, the open web, the railroads' shared clock that are owned by no one, and precisely because no single company controlled the handshake, it released the greatest wave of value we've ever seen. And there's the enclosing kind like the standard door that still only opens into someone else's house.

The encouraging news is that the backbone went the open way: neutral, owned by nobody. The seduction is to assume that settles it. It doesn't, because the platforms are fitting that open socket to their own walls. For twenty years those walls were built to keep you out; now there's a beautiful standard entrance, with an invitation for your agent to walk right in. Another brick in the wall except this one has a USB-C port in it. And the layer where the money moves hasn't been agreed at all. It's being carved up, right now, by the same handful of giants.

The question of whose noon we end up setting our clocks to is far from being settled from a full-stack perspective. Will it be a shared one the whole industry can build on, or a clock only its owner can wind?

I think it gets decided over the next eighteen months, layer by layer, by whoever chooses to own each handshake. And the marketers who grasp that this is the real contest while everyone else is at Cannes arguing about whether AI can be creative will be the ones who get a vote.

The shine was never where the power was. The power is in the plumbing. It always was.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 4 June 2026 in the Brandflow newsletter on LinkedIn.