Agentic Diplomacy
OR: What happens when the machines inherit our worst skills
There is a phrase I can’t stop turning over in my head this week, while I watch the news from Iran and feel the particular nausea of watching a thing you imagined would go badly go very, very badly.
The phrase is agentic diplomacy, and I made it up, but I think it might matter.
But first, CONTEXT!
The translation failure
On February 25th, Iran’s foreign minister said a “historic” agreement with the United States was within reach. Three days later, the US and Israel launched strikes that killed Iran’s Supreme Leader and over 1,100 civilians in six days. The stated justification has shifted multiple times. The UN’s nuclear watchdog says Iran was not close to having atomic weapons. US intelligence assessments say Iran would need until 2035 to develop an intercontinental ballistic missile.
I am not a foreign policy expert so I’m sure there are a bajillion factors at play, one of which I’m assuming is Trump’s personal diplomatic ineptitude. I am, however, a person who spent 15 years in rooms where institutions said one thing and meant another, and my job was to close that gap before it cost them something they cared about loosing.
So what I see when I look at this is not politics. I see the most catastrophic version of the problem I’ve spent my career solving: the failure to translate intent across cultural systems.
Diplomacy is cultural translation. Not metaphorically. It is the literal practice of understanding what the other side is actually communicating. Dissecting the signals underneath the words, the intent behind the posture, the meaning that lives in context (not in text), and responding in a way that is legible to a fundamentally different system of meaning.
When it works, people live. When it fails, we get war.
The brand version of the same problem
Dialed way the heck down, we have the business version, but in lieu of death and polycrisis, we have broken funnels and decreased shareholder profits. It goes a little something like this. A wellness brand says “empowerment.” Half their audience hears: they believe in me. The other half hears: they think I’m broken and need fixing. Same word. Two cultural readings. The brand doesn’t know the second one exists because everyone in the room intentionally shares the first reading in their tricked out agency deck.
Nobody dies. A campaign underperforms. An audience quietly disengages. The brand wonders why their “inclusive” messaging isn’t landing with the communities they’re trying to include.
It’s the dumbest, lowest-stakes version of the exact same skill that, at the highest level, is the difference between a diplomatic agreement and a bombing campaign.
The failure mode is always the same: you communicated for yourself instead of translating for your audience, and you didn’t know the difference because you never learned to see your own cultural logic as one system among many.
Now give that skill (or lack of) to machines
Every institution on earth in the near future will have AI agents communicating on its behalf. Not just generating marketing copy or bad AI slop product shots. They’re negotiating, responding, escalating, de-escalating, representing, exchanging. Governments. Healthcare systems. Financial institutions. Your company. My company. All of them.
And the entire industry building AI communication tools is focused on two things: what to say (identity, messaging, brand guidelines) and how to sound (tone matching, voice replication, style enforcement).
Nobody is building the layer that encodes why. The reasoning. The thing that tells a system: “When you say ‘professional’ to this audience, it signals trust. When you say it to that audience, it signals institutional distance. Here’s why, and here’s what to do instead.”
That layer doesn’t exist. Which means we are heading toward a world where AI agents communicate on behalf of institutions across cultural lines with zero cultural translation capability.
Agentic diplomacy
Now extrapolate one step further.
Soon it won’t just be a brand’s AI talking to a human. It will be a brand’s AI talking to a customer’s AI. A government’s AI talking to a citizen’s AI. A company’s procurement AI talking to a vendor’s sales AI.
Two systems. Two cultural logics. No shared framework for translating between them.
When two humans miscommunicate across cultural lines, there are repair mechanisms. The pause. The furrowed brow. The “wait, what do you mean by that?” Agents don’t have those. They’ll optimize for their respective objectives and barrel through the cultural misfire at machine speed, and neither side will know it happened until the damage is done.
A support agent tuned to be “efficient and solution-oriented” interacting with a customer agent programmed to “advocate firmly for the user” could turn a refund request into a confrontation. Who wins isn’t necessarily important because neither is wrong, but neither party has a framework for reading the other’s cultural register.
Scale that up. Make the stakes higher. Make it governments instead of brands. Make it military systems instead of customer service bots.
I’m calling this problem agentic diplomacy: the negotiation layer between AI systems operating from different cultural logics.
And right now, today, in March 2026, we are watching human diplomats — with all their lived experience, emotional intelligence, linguistic training, and capacity for nuance — fail so catastrophically at cultural translation that over a thousand civilians are dead in less than a week.
If they can’t do it, what makes us think machines will?
What I’m curious about
I’ve spent the last several months building something I’m calling the Cultural Translation Protocol. It’s an open framework and machine-readable file standard for encoding cultural reasoning into AI communication systems. It’s the layer that sits between an institution’s intent and an AI system’s output and asks: does this actually mean what you think it means to the person receiving it?
It starts with brands, because that’s my world and that’s where I can test it. But the architecture is the same regardless of whether the institution is a skincare company or a government agency.
It’s early. It’s messy. I’m not a tech person so I’m voraciously consuming and talking and reading. I don’t know if the format is right yet, or if the encoding will hold at scale, or if I’m building the wrong thing entirely. But I keep coming back to the same question: if we’re going to hand AI the job of communicating across cultural lines on behalf of institutions — and we are, whether we’re ready or not — shouldn’t someone be building the translation layer?
That’s the bet. I’d rather build it wrong and learn something than wait and find out nobody did. The cost is just too high.
What I’m reading
The Brand That Thinks — Zoe Scaman on what happens when you stop treating AI as a production tool and start treating it as something to think with. Circles the same problem from a different angle.
Strategic Sociopathy — Also Zoe. A brief lands between a news alert about wildfires and a story about a teenager who died after falling in love with a chatbot. The dissonance between commerce and reality, in one morning inbox.
Tech’s newest evil main character — Taylor Lorenz on how defense tech, AI surveillance, and far-right ideology are merging in Silicon Valley. The through-line to the Iran piece writes itself.
JAI vs. DeepSeek AI: A Historic Bias Audit — Christian Ortiz (Justice AI) running a real-time bias audit on DeepSeek. If you want to see what structural bias looks like when someone actually holds an AI system accountable, start here.
Announcing the Agent2Agent Protocol — Google’s open protocol for agent-to-agent communication, launched with 50+ enterprise partners. Read the framing: it’s all task coordination and data exchange. Cultural reasoning doesn’t appear once.
I’m Melissa. I run LOAR (Creative R&D Studio) from Lisbon, Portugal. If this piece made your brain itch, you’re my people. Reach out.