On 1 July 2026, Palantir's Alex Karp went on CNBC and said what most enterprise leaders think privately but rarely say on air: on the token-consumption model that OpenAI and Anthropic popularized, "something has gone completely wrong." The day before, Palantir had published a nine-point manifesto on "AI sovereignty" and paired it with a deal to run Nvidia's open Nemotron models inside classified, air-gapped US government environments the customer controls.

Intelligence has never been cheaper. Thinking has never been more expensive. Unfortunately, most organizations budget for the first and starve the second. AI content has never been easier to spot, as you’re probably well aware. You glance at a deck, or read the first three lines of an email, and immediately know whether it was written by a person or by AI. This will only become more obvious as models improve.

Automation. Speed. Scale. Everything luxury has always refused. And yet, AI agents may be the first technology capable of delivering what the industry has promised for decades but never quite managed to completely achieve: knowing every client as if they were your only one.

In just a few years, generative AI has moved from novelty to default front door. Large language models (LLMs) already attract close to 45 billion visits per month, the equivalent of 56% of traditional search volume, which has plateaued for years.

Most CRM and CDP projects I work on in Germany don't fail because of the technology. They fail before the first data pipeline is built. CMOs still largely lead marketing strategy within silos, brand A, channel B, campaign C. But the customer doesn't see silos. They are a business traveller booking an upgrade on Monday. A family buyer hunting a weekend sale on Friday. An insurance policyholder renewing in October. The same person. Three distinct micropersonas. Three different touchpoints. And in the database: a flat profile with a date of birth and a list of all transactions.