The Last Interface

“The next great platform will not own the destination. It will own the decision that chooses it.”

Agentic AI is shifting control from apps, search and marketplaces toward the agent that captures intent, selects outcomes and executes decisions.

These are not isolated product moves. They point to a new control layer — and the next platform war is about who controls the decision before the transaction begins.

1. 01 Interface vs. Orchestration

   Apps remain — but increasingly as callable services. The agent decides which app, which merchant, which product, which payment rail. Apple aims to become the operating system for agents; Google to move from search to execution; Amazon to own the shopping decision; Meta to own ambient context. signal: Own the orchestrator — or compete to be called by it.

2. 02 Attention vs. Authority

   The internet monetized attention. Agentic AI monetizes permission to act. That elevates identity, trust, payments, security and context. The most valuable agent may not be the smartest — it may be the one users trust to execute. signal: Authority to act outranks attention to see.

3. 03 Discovery vs. Selection

   Digital platforms monetize visibility. Agents can evaluate thousands of alternatives and return one. The economics shift from being seen to being selected — pressuring search advertising, marketplace advertising and app discovery. signal: The next discipline is Agent Optimization.

1. 01 Fewer monetizable decision points

   If agents compress search, comparison and navigation, traditional monetizable surfaces can shrink: fewer clicks, less sponsored inventory, lower discovery revenue, pressure on traffic-based monetization. signal: Compressed journeys mean fewer places to charge.

2. 02 Higher conversion, but tougher economics

   Agents can improve transaction efficiency and lower friction, raising conversion and GMV. But greater price transparency and automated comparison can pressure take rates, merchant margins, CAC and pricing power. signal: More GMV, thinner spreads.

3. 03 Margin moves toward the control layer

   The platform closest to intent and authorization can monetize through subscriptions, transaction fees, outcome-based advertising, premium routing and embedded payments. signal: Proximity to intent becomes proximity to margin.

4. 04 Capital intensity rises

   More autonomous workflows mean more inference, memory, networking and power — driving higher capex, higher depreciation, greater financing needs and higher hurdle rates. The key divergence: revenue can accelerate while ROIC deteriorates. signal: Growth and returns can pull in opposite directions.

5. 05 Valuation dispersion widens

   The market should increasingly distinguish high control plus low capital burden from high volume plus high capital burden plus weak pricing power. The transmission path: agent adoption → control shift → pricing-power shift → margin shift → ROIC dispersion → multiple dispersion. signal: Dispersion, not adoption, is the story.

1. 01 Orchestrators

   Highest strategic value sits with platforms controlling intent, context, identity, authorization and transaction. They sit closest to the economic decision and are best positioned to capture premium economics. signal: Closest to the decision, closest to the margin.

2. 02 Infrastructure With Operating Leverage

   Accelerators, memory, networking, power and data centers remain direct beneficiaries. But volume alone is not enough — watch utilization vs. capex, pricing power, cost per inference and incremental free cash flow. signal: Demand is given; unit economics are not.

3. 03 Trust and Transaction Rails

   Agents cannot execute without authorization. That elevates identity, payments, authentication, security and permissions — the toll roads of agentic commerce. signal: No execution without authorization.

4. 04 Incumbents at an Inflection

   Search engines, marketplaces and app stores retain scale. But investors should demand proof that AI expands economics rather than merely protects traffic — watch revenue per query, advertising yield, take rates, conversion and agent-driven transactions. signal: Prove expansion, not just defense.

5. 05 Execution Layers at Risk

   The greatest disintermediation risk lies where businesses do not own intent, lack proprietary data, have weak pricing power and are easily callable through APIs. Revenue may rise while margins fall. That is the trap. signal: Rising revenue, falling margin — the trap.

Key takeaways

What to watch next

• The thesis strengthens if agents generate meaningful transaction volumes, advertising moves from clicks toward outcomes, apps expose themselves as agent-callable services, payments embed into agent workflows, and inference growth comes with improving unit economics.
• The thesis weakens if agents remain primarily informational and users return to conventional interfaces to decide and transact.
• Orchestrators capture control; infrastructure captures volume; trust rails capture transaction leverage; execution layers face the greatest risk of commoditization.
• The largest revenue beneficiary may not be the largest value capturer — the critical signal is where the margin pool moves, not AI adoption itself.