Marketers Must Rethink AI Hype and Habits After Benedict Evans’ Reality Check

Benedict Evans, the technology analyst known for calling platform shifts early, brought welcomed precision to an overheated AI debate in The Knowledge Project Podcast. He argues the impact is historic, but not mystical. “This is like the biggest thing since the iPhone,” he said, “but I also think it’s only the biggest thing since the iPhone.” For communicators, that framing matters. Treat AI as a platform shift that will reshape tools and channels, not as a force that replaces strategy, audience understanding, or brand stewardship. The mandate is to build practical road maps for a fragmented, experimental landscape and resist the lure of science-fiction promises.

Adoption is slower and patchier than headlines suggest

Evans points to consistent survey patterns: “It’s like something around 10% of people are using this every day, another sort of 15 to 20% every week.” Many try it, do not get it, or return only weekly. That mirrors what many marketing teams see in pilots that never leave the lab. He pins part of the problem on product design. The blank chatbot interface forces people to invent use cases from scratch, which slows habits. The practical path for communications is to embed AI where work already happens. If a CRM button says “Draft reply” or a newsroom tool pre-fills a headline variant, usage jumps because the job to be done is obvious. Communicators should fund integrations, not sightseeing experiments, and measure adoption on daily active use inside workflows rather than demo traffic.

Consumer LLM apps feel interchangeable, so brand and distribution win

“It seems to me right now you could do like a double blind test of the same prompt given to Grok, Claude, Gemini, Mistral, DeepSeek. I bet most people wouldn’t be able to tell which is which,” Evans said. That commodity feel explains why ChatGPT leads on usage. “It’s the default. It’s the Google,” he added. For marketers, this is a positioning lesson. When functional differences are thin, brand trust, accessibility, and ecosystem distribution drive choice. Comms teams should emphasize reliability signals, enterprise assurances, and channel presence. If you are evaluating partners, prioritize stability, support, and governance. If you are marketing an AI product, invest in clear value claims and repeatable outcomes over abstract benchmarks that audiences cannot feel.

Originality still requires human judgment and feedback

“For an LLM variance is bad. Originality is a lower score,” Evans noted. Without an external signal that says different is also good, models converge on the likely. That puts a ceiling on breakthrough creative unless teams add feedback loops. Communicators can use AI to accelerate first drafts, variations, and data synthesis, then apply editorial standards and live-test with audiences to separate average from memorable. A practical rubric helps: AI drafts, humans decide, audiences validate. Evans channels the same test for his own writing: “Is this what ChatGPT would have said? Then I don’t publish it.” Make that your bar for thought leadership. If a bot could have written it, rethink the angle until it delivers distinctive insight.

Where the giants are headed and what it means for budgets

Evans believes Meta and Amazon aim to commoditize the model layer. “They want to make LLM a commodity sold at cost,” he said, with Meta differentiating above through social and ads, and Amazon differentiating below with AWS and retail ads. In practice, model access will get cheaper while value shifts to applications, data connections, identity, safety, and workflow fit. For communications leaders, budget accordingly. Spend less on raw model novelty and more on data quality, permissioning, measurement, and integration with existing stacks. Ask vendors how they will survive if the model is a commodity and how your team will benefit when the model price curve keeps dropping.

Regulate harms, not “AI”

“Talking about regulating AI as AI is the wrong level of abstraction,” Evans said. The policy focus should be use cases and harms, with clear trade-offs. Heavy, preemptive controls risk chilling startups and talent. Communicators working in regulated sectors should prepare issue maps by use case, document mitigations, and align claims to specific safeguards. Communications about AI should avoid vague reassurances. Explain what your system does, where it can fail, how you test it, and what you do when it does fail. Precision builds trust.

No obvious moats yet, so retention must be earned

Unlike operating systems, search with click feedback, or social graphs, LLMs do not obviously improve because more people use them. “There’s no apparent equivalent in LLM right now,” Evans said. Memory may create switching costs, but it is not a compounding flywheel. For communicators, this changes competitive messaging. Promise outcomes, not lock-in. Win loyalty with speed to value, task fit, and service quality. For internal programs, plan for multi-model strategies and portability. If loyalty depends on outcomes and experience, your comms should highlight measurable gains your stakeholders feel each week, not theoretical model metrics.

Reset moments favor challengers, but incumbents adapt

Evans describes a “moment of discontinuity” that resets habits and supplier choices. That threatens entrenched defaults like Google search, even if incumbents still hold advantages. The implication for marketers is to revisit channel priorities and attribution. If users route information tasks through assistants, search looks different, and the buying journey compresses. Update your content discovery strategy for LLM answers, including source transparency, structured data, and partnerships that surface your expertise inside assistant experiences. Build for the store shelf that now talks back.

The practical posture for marketers

Evans’s core message is sober and useful. Treat AI as a profound platform shift that becomes ordinary software over time. “In 10 years time it’ll just be software,” he said. The job for marketers and communicators is to turn today’s novelty into repeatable, trusted execution. That means embedding AI in workflows, competing on brand trust and outcomes, using human judgment to push beyond average, and preparing for a world where assistants mediate discovery. AI is powerful, but it is not magic. Strategy still decides who wins.