Preface | AI Wars of 2026
The central question and working thesis. A page from the AI Wars of 2026 research edition.
Preface — Where Does the Value of AI Live?
The question behind the AI wars
The obvious question in AI is: who will build the smartest model?
Does that question still matter? Yes. A real lead in reasoning, coding, multimodal understanding, planning, or scientific work can change which products people choose and which companies capture value. But model quality alone may not decide the contest.
AI products are moving past answering prompts. They now pursue goals using tools, data, applications, and human approval. That changes what competes: not just a model response, but the full system that understands a task, reaches the right context, acts with legitimate authority, recovers from failure, and keeps a person or organization in control.
That leads to the book's central question:
If people and enterprises can access several strong models, who controls the path from human intent to consequential action?
This book does not assume the answer is a platform, a model provider, an AI-native company, or any single company. It examines when different positions could create value.
The working thesis
Durable AI value may accrue to one or more control points that connect delegated intent with context, authority, execution, and economic outcomes.
An intelligence host is one potentially valuable position. It need not own every model; its value may lie in making intelligence useful inside real work, while lowering the cost and risk of switching among models, tools, and workflows. Others may capture value through intelligence, infrastructure, transactions, standards, or another control point.
The hypothesis does not assume interchangeable models, inevitable platform wins, or agents that reliably complete consequential work today. It is worth testing because companies are building products around models rather than presenting models as the complete product. Microsoft documents agents as systems that combine knowledge, actions, orchestration, models, and user experience. OpenAI describes increasingly long-horizon delegated work through Codex. Source 1 Source 2
A more useful definition of winning
The book will not ask only who has the best model or the most users. It will ask:
- Who controls access to differentiated intelligence?
- Who holds the context that makes that intelligence useful?
- Who can authorize actions safely and visibly?
- Who orchestrates models, tools, agents, workflows, and people?
- Where does the user begin when they have an objective?
- Which systems are hard to replace after deployment?
- Where do customers receive value, and where do suppliers retain margin?
Different companies may lead in different contests. A defensible final assessment may describe a balance of power rather than a single winner.
What counts as transformation
This book gives priority to operational evidence: repeated use, completed tasks, reliability and recovery, throughput, governed deployment, and workflow outcomes defined by the product. These measures test whether AI changes how a service or industry actually operates.
Financial scale is not that proof. Revenue, valuation, investment, and capital spending can show resources, demand, or commitment — but not that a workflow improved or that customers received value. Realized ROI can support a claim of sustainability when its product, costs, outcome, population, and period are clear. Expected ROI remains a scenario: some long-horizon systems, including infrastructure, commerce, and physical operations, may take years to show returns. The absence of immediate realized ROI is not failure — but an expected return is not evidence that transformation has already happened.
Source references
Source 1: Agents for Microsoft 365 Copilot
Microsoft Learn
Published: 2026
Rolling documentation covering enterprise data, actions, multi-step workflows, orchestration, models, and Microsoft 365 integration. The document can change after its verification date.
Source 2: How agents are transforming work
OpenAI
Published: 2026-06-25
OpenAI Economic Research discussion of Codex usage and delegated long-horizon tasks; task horizons are model-estimated and directional.