Everyone is watching AI make images and answer chats. Meanwhile it has quietly rerouted UPS by 100 million miles a year, cut Google's cooling energy 40%, and turned 360,000 hours of legal work into seconds. The visible tip is the demo. The submerged mass is the business — and it is now the difference between surviving and not.
The chatbot that writes an email, the app that turns a selfie into a cartoon — loud, visible, and in dollar terms, small. The value that actually moves an economy runs underneath the waterline, with no friendly interface and no screenshots. McKinsey puts $1.4–2.6 trillion a year of generative-AI value in operations and back-office1 — not in the front-of-house tools everyone talks about.
Route optimization. Fraud scoring at 143 billion transactions a year. Demand forecasting. Predictive maintenance. Claims paid in three seconds. Protein structures that won a Nobel Prize.
None of it looks like "AI" to the public. All of it is where the money is.
Adoption is no longer the question — 88% of organizations already run AI somewhere2. The question is capture, and the capture is brutally concentrated.
"It's unlikely you'll lose your job to AI. It's most likely you'll lose it to somebody who uses AI."
Jensen Huang · CEO, NVIDIA6Even the sharpest counter-argument makes the point: MIT Sloan argues ubiquitous AI confers no lasting edge7. Exactly — that's what makes it survival. AI has become a hygiene factor: the cost of staying in the game. A non-adopter doesn't lose a differentiator. It loses viability.
For twenty years the moats were slow and capital-heavy: a big data team, an expensive service operation, a knowledge base nobody maintained. AI is dissolving each of them into a commodity — and the durable advantage stops being size. It becomes speed of building.
Notice the pattern the rigorous studies keep finding: the gains are largest for novices and the least experienced.9 The old advantage — having accumulated expertise and infrastructure — is exactly what's being compressed. When what took years and a team now takes weeks and one operator, the moat is no longer what you've already built. It's how fast you compound.
The value lands in operations — the systems with no interface, that never trend, that quietly decide whether a business wins. A sample of what's already in production:
The same technology that produced JPMorgan's 360,000 saved hours is the one stalling in 95% of pilots. The difference is where the AI lives.
Klarna is the whole lesson in one company: the "replace 700 agents" framing made a headline, then broke on the complex cases, and the fix was a governed human-in-the-loop model.22 The 95% isn't evidence against AI — it's the price of doing it top-down and ungoverned.
This is the skill your people have to learn. Not "prompting" — the shift from being an executor of tasks to being a solver and orchestrator who directs a portfolio of AI toward an outcome. Microsoft calls the new role "agent boss": everyone, from intern to C-suite, overseeing their own constellation of agents.23
"AI won't replace humans — but humans with AI will replace humans without AI."
Karim Lakhani · Harvard Business School / HBR26And ordinary employees become the builders. At Moderna, staff created 750 custom AI tools in two months — 40% of users built their own.27 No vendor ships 750 internal tools in two months. The people who own the workflows do — because they're the only ones who know where the friction is.
Twenty-five years in digital strategy and media — and a lot of it in games. I directed marketing at Hoplon and at Level Up Games: a 32-person team, esports, a 24-hour streaming studio we built with our own people, influencer programs, GDC. I know this industry from the inside.
In 2017 I watched my own value start to commoditize — the rise of the "button-pusher." So I pivoted into AI and predictive systems before it was fashionable. I was a Chief Marketing Technology Officer in 2021, before the generative-AI wave — the title just named what I'd already been for two decades: the bridge between marketing and technology.
My through-line never changed: marketing as a system you operate, not a cosmetic layer you apply. Today I'm Director of Strategy & Media at a brand agency — where I don't just advise, I build the systems the team uses and mentor them to build their own. I'm the guinea pig.
I'm the guinea pig. My own marketing operation is AI Native in the literal sense — systems I built, running today, on infrastructure I control. Not a demo. A workday.
There's a name for it, and it's twenty years old. After Deep Blue beat him in 1997, Garry Kasparov invented "advanced chess" — and found that the strongest players on Earth weren't grandmasters or engines. They were humans who orchestrated their machines with a better process.28 The best meta-analysis to date confirms the sharp version: the orchestrated human beats both the AI and the unaided expert — and everyone who just turns the tool on and trusts it does worse.29
This is the real frontier of the whole shift — not a smarter machine, but a larger human. The augmented professional. I'm not ahead of it; I'm just practicing it out loud, on myself, every day — so that when I teach your team, I'm handing them something I've already lived.
Two design rules cut across everything I build — and they matter more for a games publisher than almost anyone, because your community already drew a hard line at generative AI in anything player-facing.
Concrete, buildable systems for the exact teams you named — creative, PR, community, social, influencer, web, CRM. Every one faces inward. None puts AI in front of a player.
Not a training course people forget. Up to five representatives — one per discipline, the person who owns the workflow and will own the system.
The deliverable isn't the systems. It's a team that no longer waits for a vendor.
The deliverable was never the systems. It's a team that no longer waits — and a company that stopped experimenting and started compounding.
The gap compounds every quarter. For a publisher whose community already drew its line on AI, the winning move is the opposite of the mistake everyone fears: embed AI in the operation, govern it, keep it invisible to players, and teach your people to build. If this is useful, I'd be glad to run a working session with your marketing leads — remote or in Stockholm.
Notes: Klarna's "700 agents" reflects hiring avoided during growth, not layoffs; the company later re-added human agents for complex cases — cited here as evidence that AI must be governed, not as a headcount play. WEF's "39% of skills" (skill-sets transformed) and LinkedIn's "70%" (skills used in a job) measure different things and are presented as two lenses. Consultancy macro-figures (McKinsey, BCG, PwC) are estimates, cited as directional. Every linked source was checked at time of writing.