Technical Context
I'm seeing the same pattern more and more often: a company announces an "AI transformation," launches a couple of pilots, buys API access to a model, builds an internal chatbot, and six months later, the whole thing gets bogged down in approvals and quietly drowns in legacy. Against this backdrop, Block's approach is compelling not for its hype, but for its rigorous execution.
Looking at the public facts, Block's initiative wasn't a one-off "let's tack on an LLM" project. They went for a complete overhaul: infrastructure, cloud, analytics, MLOps, internal AI agents, and a new operating model for the engineering team. In other words, it wasn't a cosmetic change but a fundamental shift in their operations.
I've looked through the available materials on the Block case, and the most interesting part isn't the Goose agent or the slick AI manifesto. What caught my eye was that they didn't try to slap AI onto a messy foundation. They first created an environment where an AI architecture could exist without workarounds.
Here's where things usually break:
- Data is scattered across systems with no real ownership.
- Access, compliance, and security kill speed long before production.
- Teams build ten disconnected AI features without a unified architecture for AI solutions.
- Leadership expects magic from a model, when the real problem lies in processes and accountability.
Against this backdrop, talk of "just gradually adapting" sounds nice, but in the enterprise world, it's often just a way to feel better. Gradualism works when a company already has a solid foundation. Without one, it just stretches the chaos out over time.
The contrast with Klarna is telling. There was a lot of talk about their drastic measures, layoffs, and big bet on AI, but public noise and sustainable business transformation are not the same thing. If you have to rehire people or rebuild processes from scratch later, the problem wasn't the pace but the flawed change model.
Impact on Business and Automation
I would stop looking at AI transformation as a project with an end date. That framework is obsolete. The winners today aren't those who have "transformed," but those who have learned to operate in a state of endless adoption, where AI-powered automation becomes a continuous capability, not a special project for a single fiscal year.
For a business, this changes everything: budgeting, team roles, tech stack, IT priorities, and even hiring. You don't just need analysts and developers; you need people who can build an AI architecture for a specific scope: deciding where to keep a human in the loop, where to grant an agent autonomy, and where to mitigate risk with rules and audits.
Companies that are willing to cut unnecessary approvals, consolidate data, and quickly push working scenarios to production are the ones that win. Those who pretend that implementing AI is like a standard ERP upgrade lose. The turbulence is too high, and the tools themselves are changing too fast.
I see this in Nahornyi AI Lab's client cases as well. When a business comes with a request to "build us one AI agent," it almost always uncovers a deeper challenge: re-engineering data flows, redefining ownership, and linking CRM, helpdesk, documents, and internal databases before even starting the AI integration. Otherwise, the agent just automates the chaos.
Therefore, I would formulate the "Block method" more simply. You don't necessarily have to fire all the saboteurs, as people love to say in discussions. But you definitely need to stop trying to persuade a system that is structurally resistant to change. Sometimes, it's cheaper to radically rebuild a process than to patch up the old one for years.
If you need to move beyond the hype and actually implement AI automation, create an AI agent for your process, or build a working architecture for AI solutions, this is what I do. I'm Vadim Nahornyi of Nahornyi AI Lab. If you want to discuss your case, order AI automation, or n8n automation for a business task, contact me, and we'll dive into the specifics.