The Human Agency Imperative | AI Wars of 2026
Voluntary judgment, value, trust, and agency. Source-backed analysis from AI Wars of 2026.
Part 0 — The Wars
The Human Agency Imperative
AI competition does not end when a capable system becomes available. A model, agent, or platform becomes strategically consequential only when people decide it is worth using in the work and choices that matter to them.
The human side of the contest is not a battle to own or manipulate people. It is the challenge of earning voluntary judgment: attention, perceived value, trust, autonomy, and the willingness to make an AI system part of a repeated practice.
“... changing behavior is just much harder than the tech nerds realize.”
— Sam Altman, in conversation with David Senra, August 2026 Source 1
Big Idea
There is a second contest: earning people's trust. People only keep using a system they find valuable, trustworthy, and fit for their real work.
This is a strategic condition, not a claim that every person, organization, or market responds in the same way. A system can be widely available and still fail to become a durable practice. It can be technically capable yet fail to earn the trust, perceived value, or fit that continued use requires.
The human side of the contest
Technology-adoption research distinguishes access to a system from the conditions that influence whether people choose to use it. AIRS extends the UTAUT and UTAUT2 research tradition into AI adoption, examining readiness and behavioral intention in a U.S. academic and professional sample. Source 2 Source 3 Source 4
That distinction matters for AI strategy. Distribution can create exposure. Training can improve a person's ability to use a system. Neither automatically establishes that the system feels worthwhile in a particular workflow, that its authority is acceptable, or that the person will return to it after a first encounter.
A strategic failure mode
A common response to weak adoption is to assume a skill gap: provide more training, add more features, or increase access. AIRS suggests a more demanding diagnosis. In the study, Facilitating Conditions did not independently predict behavioral intention once the other readiness dimensions were considered. Those conditions included access to training or tutorials and tool compatibility. Perceived price value was the robust reported predictor. Source 4
This does not mean that training never helps. It means a perceived-value problem calls for a different question first: what makes continued use worth the time, effort, or cost for this person in this workflow? A strategy that cannot answer that question may mistake exposure for adoption.
Reported use and perceived value
Bitkom's September survey release illustrates why use and perceived value need separate questions. In a telephone survey of 603 German companies with at least 20 employees, conducted in calendar weeks 28 to 33 of 2026, 57% reported using AI. Among companies using AI, 59% said they were not exploiting its potential at all. That is their assessment of unrealized potential, not a finding that AI produced no benefit or a calculation of financial return. Source 5
This company-level survey does not replicate AIRS, measure individual readiness, or explain why a particular person continues using a system. It shows why an adoption headline cannot answer the value question on its own. Testing that relationship requires evidence linking the same people's judgments to their subsequent behavior.
Agency is part of the prize
The strongest route to repeated use is not necessarily the one that removes the most human judgment. In consequential settings, people and organizations need to understand four things: what an AI system can access, what it is permitted to do, when it can commit them to an outcome, and how they can intervene. These are conditions of legitimate delegated action, not obstacles outside the product. Source 6 Source 7
This changes the strategic question. The contest is not only for attention or interface reach. It is also for the trust and agency that let people choose an AI system repeatedly without surrendering meaningful control.
How this connects to the wars
Chapter 0, The AI Agent Contest, maps the strategic positions around intelligence, context, authority, execution, and economic capture. This chapter adds the human condition that connects those positions to outcomes. An asset or control point matters only when people and organizations can decide that it creates value in practice.
The AIRS page contains the study's bounded evidence, predictor charts, and an optional self-assessment. It does not turn the dissertation into a universal adoption model or a claim that the AIRS Enterprise assessment itself changes behavior.
Key Question
Which AI systems earn voluntary, repeated use by creating value people can recognize, trust, and retain agency over—and what evidence would show that this judgment becomes a durable outcome?
Sources
Source references
Source 1: Sam Altman, OpenAI
David Senra
Published: 2026-08-23
Transcript of Sam Altman's August 2026 conversation with David Senra. Altman describes economic inertia, ingrained habits, and the difficulty of changing behavior even when technology is available. It is an executive's personal account, not independent evidence of general adoption behavior.
Source 2: User Acceptance of Information Technology: Toward a Unified View
MIS Quarterly
Published: 2003
Venkatesh, Morris, Davis, and Davis introduced UTAUT as a unified model of technology acceptance and use. It is a theoretical foundation for AIRS, not evidence for AI-specific adoption outcomes.
Source 3: Consumer Acceptance and Use of Information Technology: Extending the Unified Theory of Acceptance and Use of Technology
MIS Quarterly
Published: 2012
Venkatesh, Thong, and Xu extended UTAUT to UTAUT2. AIRS uses its constructs as a theoretical foundation while testing an AI-specific instrument in a separate study.
Source 4: Artificial Intelligence Readiness Scale: Extending UTAUT2 for Enterprise AI Adoption
Touro University Worldwide / Fabio Correa
Published: 2026-04
Doctoral dissertation reporting validation of AIRS-16 using a U.S. academic and professional sample (N = 523). Its cross-sectional findings support sample-bounded associations with behavioral intention, not universal causal effects, realized adoption, productivity, retention, or business value.
Source 5: Erstmals nutzt die Mehrheit der Unternehmen KI
Bitkom e. V.
Published: 2026-09-14
Bitkom reports that 57% of surveyed German companies now use AI, against 36% a year earlier and 20% two years earlier, and that 4% say AI is not a topic for them. Its methodology note states that Bitkom Research telephoned 603 companies with 20 or more employees in Germany, in calendar weeks 28 to 33 of 2026, and describes the survey as representative of the economy as a whole. The same instrument reports that no company using AI says it fully exploits the potential, 3% say they exploit it strongly, 28% little, and 59% 'ueberhaupt nicht' - not at all. The measure is self-assessed at company level and 'AI use' is not defined on the page, so it establishes breadth of use rather than depth, outcome, retention, or value. The fieldwork period precedes publication by roughly six to eight weeks and must not be read as a measurement of the publication date.
Source 6: Model Context Protocol Specification
Model Context Protocol
Published: 2026-07-28
Current specification for a protocol connecting LLM applications with external data sources and tools through hosts, clients, and servers. The 2026 revision uses stateless, self-contained requests and adds opt-in extensions; standardized integration does not make prompts, memory, identity, authorization, or workflows fully portable.
Source 7: Artificial Intelligence Risk Management Framework (AI RMF 1.0)
National Institute of Standards and Technology
Published: 2023-01-26
Voluntary, use-case-agnostic framework organized around GOVERN, MAP, MEASURE, and MANAGE. It does not certify an AI system as safe or prescribe universal approval thresholds.