Raja Rajamannar, former Global CMO of Mastercard, observed: everyone in marketing has access to identical AI tools, uses similar prompts and consequently produces comparable output. He called it a “massive Sea of Sameness”. In day-to-day agency work that finding is immediately recognisable.
The uncomfortable part
Rajamannar sharpened his analysis: prompt engineering — still celebrated in 2023 as a differentiator — loses its edge as soon as AI models can generate better prompts themselves. That shifts the competitive advantage fundamentally. To where exactly?
AI assistants generate under 0.5 % of all visits — but deliver around 11 % of B2C revenue. The conversion rate is three times higher than paid search.
The data makes it tangible: according to Neil Patel (28-country data set), 51 % of traffic already comes from non-human sources. The conclusion is uncomfortable: standard output gets cheaper, while the value of brands that AI systems recommend rises exponentially. Uniformity gets cheaper, differentiation more valuable — yet investment still follows the opposite logic.
A word nobody understood in 2014
SlopeLift positioned itself in 2014 as a “workshop” in performance marketing — an approach the industry judged at the time as charmingly old-fashioned. The focus was on granular campaign structure and individual ad copy instead of mass replication. In 2026 that concept has become a strategic competitive position.
When machines take over the average, what is left as the distinguishing feature is: judgement, contextual knowledge, intuition and decisions no model can make — because it does not know the client relationship.
A concrete example: one account reliably shows a ROAS collapse in calendar week 12. An agent would cut the budget — and would be wrong every time. The collapse is payday week in the target group; behaviour normalises a week later. That knowledge sits with one person who has watched the account for three years. It does not exist in data sets.
What this means for AI leaders
The central industry question has shifted: from “how do we produce more, faster, cheaper?” to “what can we do that a competent model with competent input cannot copy by Friday?”
The answer lies in three dimensions:
- Proprietary data: signals competitors cannot buy.
- Talent: people who steer the system and supply the context — the skills gap that comes out of this is described by AI-Sorting.
- Leadership: a willingness to make real bets instead of imitating competitors’ decisions.
Technology becomes a commodity. These three factors do not. The efficiency of AI sharpens the craft — agents give back time that can flow into questions for which no prompt exists.
A critical question for everyone who owns AI in a company: how much of last quarter’s output could a competitor with identical tools and slightly better prompts have produced just as well?
That percentage is the size of your own sea. Everything else is the island.