[Two Cents #75] “Flights of Thought” on Consumer + AI — Part 1: Fundamental Shifts
Introduction
It’s becoming clear that the market’s “readiness” for Consumer AI has crossed a tipping point.
What we need now is to get far more concrete about how AI-driven market change will unfold—the direction, the mechanisms, and the implications for industry structure, competitive dynamics, and the economics between participants.
For founders, the job is to identify those opportunities a little earlier and move first. For investors, the job is to recognize those early moves quickly and support them aggressively.
This series—my “Flights of Thought”—is an attempt to share how I’m thinking through what will happen, what it will unlock, and what kinds of ideas are likely to matter.
Fundamental Shifts
The first thing to pay attention to is that AI is driving structural change in the internet and mobile-era playbooks we’ve all internalized.
This isn’t just about new startups launching new products. It’s about the counterparties startups interact with—platforms, distribution channels, partners, competitors—shifting underneath them. When that happens, the economics of competition and transaction change as well. The relationship map changes, and with it, the revenue model and unit economics a new company can realistically build on.
So before debating specific “startup opportunities,” we need to understand the foundational shift in market structure that’s now underway.
Search and SEO
It’s already obvious that Google’s search-driven SEO structure is changing at the root.
Yes—search via chatbots like ChatGPT is still tiny relative to Google (on the order of a fraction of a percent). But even just adding AI Overviews has materially altered organic traffic flows to blue links—by something like 30–60% in many cases. If conversational search (e.g., AI Mode) becomes the default interaction pattern, it’s not hard to imagine the first real cracks forming in Google’s historically unassailable position.
In commerce—where visibility is oxygen—we’re already seeing early movement toward AI SEO, and the shift from “SEO” to what many now call “GEO” is increasingly inevitable.
More importantly, the Agent-to-Agent service architecture that’s now starting to spread will reshape discovery, selection, and transaction across most online commerce. I’ll go deeper on Agent-driven markets in later posts; here I’ll focus on the shape of the shift.
The biggest change is simple: the “actor” executing a transaction moves from humans to agents, turning many interactions into machine-to-machine (M2M) flows. Once that happens, every business model built on the assumption that “a human will browse, compare, and click” is forced to change.
Google Search’s original logic is: show a ranked list of relevant links, assume humans will evaluate them, and compete to be the link humans click. SEO is the arms race to win that ranking and that click.
Two things have now changed:
The decision-making layer for “which result matters” is shifting from human evaluation to AI mediation.
The place where discovery happens is shifting away from Google toward other surfaces—LLM chat, AI-native search products, and agent interfaces.
That is the core of the SEO → GEO transition.
At a deeper level:
Information retrieval is changing from “collect fragments via links and synthesize yourself” to “receive a synthesized answer (and, when needed, deeper research and analysis) as the default interaction.”
Even more significantly, the highest-value part of search—the path from purchase intent to purchase decision—is increasingly likely to migrate away from search as we’ve known it. Instead of users actively searching for the best product, AI and agents will increasingly find and propose the best option on the user’s behalf.
That’s why this isn’t just an SEO story. It’s a search business model story—and potentially a major re-platforming of the entire purchase funnel.
Commerce and Marketplaces
The same pattern is playing out in commerce. With “agent commerce,” the entity converting intent into action shifts from human to agent, and the logic of discovery and selection—and the stakeholder map—changes fundamentally.
There are two structural shifts to watch.
1) What happens when incumbents absorb agents.
The early example is Amazon’s “Buy for Me”-style agent shopping direction: consumers can use Amazon as the default destination, but access third-party inventory and complete the full purchase cycle. In that world, Amazon becomes an even stronger front door to commerce—while also creating a strategic dilemma: if third-party inventory is reachable through Amazon, then Amazon’s moat in “exclusive destination value” can weaken at the margin.
As agent commerce becomes real, it will impact everything tied to the purchase cycle: product pages, keyword ads, brand marketing, and retail search optimization. And this matters because retail ads are not a side business—Amazon’s shopping ad engine is enormous and contributes disproportionately to profit. This structural shift is a real threat to incumbents, but it will also create large whitespace for startups (the exact shapes are still emerging).
2) The shopping “surface” itself is moving.
The next shift is commerce moving into non-commerce platforms—LLM chat products and AI-native apps—where the entire loop from discovery to checkout is completed inside the AI interface.
In the classic flow, shopping is distributed across multiple layers: search (Google/Amazon/comparison sites), product pages (commerce platforms), browsing and comparison, checkout (wallets/PSPs/platform checkout).
In the AI-native flow, that whole loop can collapse into one interface: search → discovery → decision → checkout, happening inside ChatGPT/Gemini/Perplexity or a new AI app layer.
Greg Brockman’s early ChatGPT-4 demo—planning a meal, ordering groceries via Instacart, reviewing a cart, then completing checkout—was a clear preview: B2C platforms risk being pushed into the role of backend execution layers behind the AI front-end. Commerce is now where that shift begins to feel tangible.
Zooming out: this doesn’t stop at commerce. It extends to marketplaces and any platform that intermediates discovery, connection, and transaction. Generalized further, it implies that many B2C platform playbooks built over the last 30 years—from search to commerce to marketplaces—could see their operating models fundamentally rewritten.
Compared with that, the web-to-mobile transition looks relatively incremental: same democratized distribution structure, new form factor, richer modalities (always connected, location, camera). The AI transition is more profound—it replaces the division of labor among market participants with a different process and a different structure.
Software
Software and SaaS won’t be exempt.
Coding agents and “vibe coding” are already changing what software creation even means. And once the production function changes, the SaaS market structure has to change with it.
The emergence of AI-native startups that hit scale with radically different headcount, timelines, and go-to-market dynamics is a signal that the old SaaS playbook—ARR milestones, T2D3, conventional LTV/CAC heuristics, net dollar retention as the north star—won’t map cleanly onto what’s coming.
We’re watching a new playbook form in real time: sell work, not tools, price outcomes, rethink what it means to “build,” “serve,” and “use” software.
At an even deeper level, it’s fair to ask whether the term “SaaS” will still be the right abstraction a decade from now—or whether the concept of “software” itself shifts.
(Aside: one way to think about “software” is that it has historically been humans doing the work to translate intent into the limited language computers can understand—machine code to assembly to higher-level languages to scripts. We’re now reaching a stage where computers can increasingly understand work described in natural language and execute it. If the worker becomes far smarter—and has rich context about you, your data, your preferences—it starts to resemble a “great operator” or even an old-school butler: you don’t need to specify everything precisely; you express intent, and the system figures out execution.)
Net-net: across categories, the playbooks we’ve relied on for the last 20 years are being rewritten.
The specific form differs by sector, but the common thread is: the old playbooks are not valid in an AI-native economy. We have to write new ones.
Key Takeaways
Over the last 10 years (mobile and SaaS), and arguably the last 20+ years (web), we treated many industry playbooks as “default truths.” Most of them are no longer safe assumptions.
Trying to grow by layering incremental AI features onto existing products—while staying inside the same old distribution, monetization, and organizational frameworks—will increasingly fail.
One thing that has frustrated me over the past 2–3 years is how many early consumer startups were still chasing problem statements that looked structurally similar to what we saw over the prior decade. My hope is that the “AI opportunity set” pushes more founders to look for opportunities in fundamentally new frames.
In this environment, what matters is less “what happens if we tweak the old system?” and more:
What structural shifts are happening (or will happen)?
What are the first-order and second-order effects?
What new solutions become necessary—or newly possible—because of those shifts?
That kind of first-principles thinking is not optional. It’s required.
Across every category, we should question the default assumptions, analyze the underlying structure, and move from “finding” incremental value to building new value from the foundations up.
Aside 1: Incumbents Won’t Simply Be Disrupted
One important difference versus prior platform shifts is that incumbents may not just be targets of disruption—they may remain meaningful actors.
We already saw this in enterprise AI: Microsoft, Adobe, Oracle, Databricks, and others have adapted faster than many expected.
Why?
At the macro level, incumbents have learned from previous cycles, and their tech literacy is materially higher than it was in earlier transitions.
More concretely, in enterprise AI, incumbents often control the customer’s data pipelines and workflows. That creates a real data and distribution moat versus new entrants.
That said, the consumer market is different. Consumer opportunity is more tightly tied to behavioral shifts enabled by tech shifts, and consumer lock-in to legacy data infrastructure is weaker. In newly created consumer markets, incumbents are often less equipped to adapt—and new entrants are more likely to define the category.
Aside 2: AI Data Center Build-Out Feels Familiar
The current “$1T AI data center” conversation looks structurally similar to the telco infrastructure cycle from around 2000—fiber build-out, data centers, overhang, then consolidation.
The scale is larger—maybe 10–100x—partly explained by inflation and partly by compute intensity. But the pattern rhymes.
It wouldn’t surprise me if, 20 years out, we see echoes like: fiber glut → power glut, and telco consolidation → AI data center consolidation.
That doesn’t mean the opportunities and risks in this build-out are irrelevant. It just means: this infrastructure story is important—but it is not “the entire world.” It’s a major layer, not the whole stack.
Call for Startups
The purpose of sharing this thinking is straightforward. As an early-stage investor focused on Consumer + AI, I hope this series helps existing startups better leverage AI-driven shifts—and helps new founders reduce trial-and-error as they search for meaningful opportunities.
In that sense, this is Two Cents’ version of a Call for Startups.
If you are an early-stage founder or startup in Consumer + AI and believe you are onto something, my inbox is always open. Feel free to reach out via DM or email:
hur at hanriverpartners dot com