I spent a day at SlatorCon London listening for one thing. Here’s what I heard.
Yesterday I posted about attending SlatorCon London with Elaine Cowley, co-founder of Sound AiSleep, and said I was curious about one thing: how the room would define “build.” Because in the build vs buy debate that enterprise AI strategy is having right now, where you draw that line changes everything.
I got my answer several times over. I also got one I wasn’t really expecting.
The CEO who was coding at 2:30am
Georg Ell, the CEO of Phrase, one of the world’s leading language technology platforms, used by enterprises across the globe to manage localisation at scale. In his segment he told the room at SlatorCon that he had never written a line of code before February of this year. He is now the highest token spender at his own company. He was coding until 2:30am the night before the conference and still had a model running on his laptop on stage.
His definition of build, in his own words: “I’m building stuff for which there’s no one else to build. I’m building stuff that makes my life easier, allows me to run the company more effectively, for which there is no software.”
The CEO of a platform that thousands of enterprises rely on to manage their language operations is teaching himself to code at 2:30am because what he actually needs does not exist off the shelf. Nobody else can build it because nobody else is him.
That is not a technology story. That is a human story about what happens when the tools run out before the problem does.
The chess player, the iPhone, and the banana ketchup
Stéphane Cinguino, Chief AI and Technology Officer at Acolad, opened with three analogies that said the same thing three different ways.
When Kasparov lost to Deep Blue in 1997, most people assumed it was the end of human chess. What actually happened was a new generation of players started training with the machine, saw the board completely differently, and raised the entire field to a level that neither the human nor the machine could have reached alone. The machine did not replace the human. It changed what the human needed to know.
When Steve Jobs revealed the iPhone, the revolution was not the phone. It was the collapsing of multiple categories into one platform. Stéphane’s argument: AI is doing the same thing to the entire content lifecycle right now. Creation, adaptation, review, brand voice, multimedia, and localisation are all being reshaped simultaneously.
Then came the banana ketchup. In the Philippines, when tomatoes were not available, they made ketchup from bananas. Same job. Different ingredients. Different process. Similar outcome. His point: the outcome matters, but the recipe can change. The localisation industry has been so attached to its tomato-based recipe of segments, translation memory, and post-editing, that it has stopped asking whether the recipe still serves the outcome.
The human judgement that decides when to change the recipe, and what to change it to, is not something the model provides. That is AI decision-making at its most fundamental, and it belongs to the person, not the platform.
Recursive AI and the regulated market problem
The investors session surfaced a dynamic that did not get enough airtime in the main programme, and it is the one with the longest commercial tail.
One investor described what they called recursive AI: a combination of AI and human intervention, cycling through each other iteratively rather than sequentially. Not a handoff from human to machine, a loop. The human shapes the output, the model learns from that shaping, the human refines again. That loop, when it works, compounds. When it does not have a human with clear intent at the centre of it, it drifts.
The regulated market dimension is equally significant. Governments in France, Germany, and across the Middle East are increasingly demanding that AI tasks be carried out within their own jurisdictions. Data sovereignty is moving from a compliance checkbox to a strategic requirement. One investor noted that businesses in Germany have already banned Microsoft. The French market alone is absorbing two to three billion dollars in AI investment this year, driven not by commercial preference but by the fact that French companies and government institutions have no choice but to keep their data and their AI operations on French soil.
The implication for enterprise AI governance is direct: the build vs buy question is not only about capability and cost. In regulated markets, it is about where the human decision-making sits, who owns the outcome, and whether the governance structure around the AI can be demonstrated to a regulator. Speed is rewarded in the US. In Europe and the Middle East, compliance, AI oversight, and human governance are the price of entry.
The pattern across every session
Once you hear that framing, you hear it everywhere. A speaker from Coca-Cola’s localisation team pointed out that AI language models are English-dominant, and that translation is not the same as creating content in a specific language. The human cultural layer, the local knowledge, the understanding of how a market actually thinks and behaves is irreplaceable, and the entire room knew it. The metrics just do not capture it yet.
In the data session, the closing observation was this: the underestimated factor will always be the importance of humans inside the process. Human judgement and human input are not optional extras. They are the thing the model is trying to serve.
Three sessions, four industries, the same gap named every time.
The thing nobody said
What struck me was not what was said. It was what was not.
Every session identified the human as the critical variable. Not one session addressed how to build around that human systematically. How to define what that human needs from AI before the building starts. How to make sure the system serves the person rather than the other way around. How to make that human intent legible to the model across every tool, every agent, every platform.
That gap is not an oversight. It is the missing layer in almost every digital transformation I have encountered across thirty years of international project delivery.
It is why I built contAIn™. Not because I had a business idea, but because the methodology to direct AI around the human operating it did not exist, and I needed it.
Georg Ell built at 2:30am because the software he needed did not exist. Coco and Elaine built Sound AiSleep, an app that clones a parent’s voice to narrate personalised bedtime stories for children, because the technology existed but nobody had built it around the thing that actually matters to a child at bedtime, which is hearing the voice of the person they trust. Three builders. The same starting point every time: the tool existed, but the human layer around it did not.
The build vs buy question is real and worth asking. But there is a question that comes before it, and I did not hear it asked out loud once today.
Everyone in that room said “human in the loop.” Not one person questioned whether the loop was the right place for the human to be.
The human is not in the loop. The human is above it.
configure YOUR system. contAIn the chaos. control YOUR outcome.
This article was originally published on Medium.