The software goes live on Monday.
The technical team has tested the integration. The permissions are configured. The vendor has delivered training. The policy has been signed. The dashboard is clean and the procurement process is finally over.
Everything that can be checked in a deployment spreadsheet is green.
Then someone asks a different question.
When the system produces an answer that looks right, who knows when not to trust it?
The room gets quieter.
Who has enough domain knowledge to detect a subtle error? Who understands what data the system never saw? Who can challenge a recommendation without slowing the team so much that challenge becomes socially expensive? Who owns the final decision? What happens when a junior employee notices a problem that a senior manager has already accepted? What parts of the work should still be learned without assistance because future judgment depends on having practiced them?
The system is ready.
Human readiness is a different problem.
Definition
The Institute of Inner Technology uses human readiness for the AI age to describe the degree to which people and institutions possess the capacities, competencies, practices and decision conditions needed to use artificial intelligence with informed judgment, preserved agency, continuing learning and meaningful responsibility.
This is an Institute synthesis.
It is not a claim that there is one measurable readiness trait inside a person.
Readiness is relational.
A person can be ready for one system, task and level of consequence while being unready for another. A team can understand a model technically while lacking a culture in which it can be questioned. An organization can have strong governance documents while leaving frontline staff too little time to verify outputs. A student can be highly fluent with generative tools while becoming less practiced at the forms of reasoning the curriculum is supposed to develop.
Readiness therefore sits at the intersection of person, task, tool, institution and consequence.
Readiness is not enthusiasm
Technology adoption often treats willingness to use a system as a sign of readiness.
It is not the same thing.
A person can be enthusiastic and poorly calibrated. They can be cautious and highly competent. They can use a system frequently without understanding when its output is unreliable. They can know its limitations and still lack the authority to act on that knowledge.
Readiness includes the ability to engage.
It also includes the ability to hesitate.
The relevant question is not “Are people comfortable with AI?”
It is “Can they use this form of external intelligence in ways that preserve the human capacities and responsibilities the situation still requires?”
AI literacy is one layer of readiness
The European Union now places AI literacy inside the legal environment of deployment.
In July 2026, Regulation (EU) 2026/1744 revised Article 4 of the AI Act. Providers and deployers must take measures to support development of AI literacy among staff and other people operating and using AI systems on their behalf, taking into account technical knowledge, experience, education, training, context and the people affected. The revised provision clarifies that providers and deployers are not required to guarantee a particular level of AI literacy in every individual.
This matters.
People need to understand what kind of system they are dealing with, what it can and cannot reasonably be expected to do, and what risks or consequences attend its use.
But literacy does not exhaust readiness.
A radiologist can be literate about an AI tool and still face workload pressures that make independent review difficult. A teacher can understand generative AI and still be uncertain which student tasks should remain unaided for learning reasons. A civil servant can know that an automated recommendation requires human oversight while having no practical power to alter it. A creative professional can understand model limitations and still discover that habitual delegation has changed the way they originate ideas.
Knowledge about AI is necessary in many contexts.
Readiness concerns what a person and institution can do with that knowledge under real conditions.
UNESCO makes human agency part of AI competence
UNESCO's AI competency frameworks for students and teachers, launched in 2024, are useful because they refuse to define AI competence as technical operation alone.
The student framework includes a human-centred mindset, ethics, AI techniques and applications, and AI system design. The teacher framework adds AI pedagogy and professional development. Across these frameworks, human agency, accountability and responsible use are explicit concerns.
These are policy frameworks, not causal evidence that a particular competency programme produces responsible behavior.
Still, they show an important shift in institutional thinking.
The educational question is no longer merely whether people can operate AI.
It is how they can remain agents while doing so.
Capability changes faster than static readiness programmes
One reason readiness is difficult is that the object of readiness keeps moving.
The OECD's 2025 AI Capability Indicators were developed to compare AI performance with human abilities across nine domains: language, social interaction, problem solving, creativity, metacognition and critical thinking, knowledge, learning and memory, vision, manipulation and robotic intelligence.
The framework is deliberately designed for updating.
That matters because a competence programme based on what AI could not do two years ago may become obsolete while the programme is still being delivered.
Readiness cannot therefore mean memorizing a permanent list of machine weaknesses.
It requires calibration as a continuing practice.
What can this system do now?
On this task?
With what evidence?
Under what conditions?
What remains difficult for it?
Which human capabilities become more important because the technical boundary has moved?
Occupational exposure is not the same as human readiness
In 2026, the OECD mapped its AI Capability Indicators to occupational requirements.
The resulting exposure measure found AI capabilities closest to occupations dominated by routine information processing, administrative work and codifiable tasks, and furthest from occupations requiring contextual judgment, interpersonal understanding, complex decision making and responsibility.
The authors make a crucial qualification: technical exposure does not mechanically determine labor-market outcomes. Adoption, regulation, organizational change and social choice matter.
Human readiness lives inside that qualification.
Two organizations can deploy comparable tools and produce very different human consequences.
One may use AI to remove low-value repetition while preserving apprenticeship, judgment and meaningful decision rights. Another may remove the very intermediate tasks through which novices once learned enough to become experts. One may create time for deeper contact with clients. Another may use the efficiency gain to increase workload until verification becomes ceremonial.
The capability of the tool does not determine the developmental architecture around it.
Six dimensions of readiness
The Institute treats readiness as a portfolio rather than a score.
The following dimensions are synthesis categories. They organize existing evidence; they are not proposed as a validated psychometric scale.
1. Technical and contextual literacy
People need sufficient understanding of the system, its purpose, data context, limits, failure modes and the consequences of its use.
The level required depends on role.
A clinician, developer, manager and patient do not need identical knowledge. Readiness is role-sensitive.
2. Domain judgment
Fluent AI output can reduce the visible cues that normally signal uncertainty.
Domain knowledge is therefore not made irrelevant by better generation. It may become more important for evaluating what generation produces.
Judgment includes knowing which distinctions matter, which exceptions are consequential and when a plausible answer is not enough.
3. Metacognitive calibration
Risko and Gilbert's cognitive-offloading framework emphasizes that decisions to use external support are partly influenced by people's estimates of their own ability. Those estimates can be wrong.
Readiness therefore includes some capacity to calibrate confidence.
When should I solve this myself?
When should I use the tool?
When should I verify?
When should I ask someone else?
The answer cannot be a blanket preference for self-reliance or automation.
4. Agency and decision rights
A person cannot exercise responsible judgment if the system around them makes disagreement nominal.
Agency is not only an inner feeling of autonomy. It includes actual authority, available alternatives, escalation routes and enough time to act.
A “human in the loop” who is expected to approve hundreds of outputs at speed may be formally present and functionally absent.
5. Learning continuity
Soderstrom and Bjork's distinction between learning and performance is particularly important in AI-supported work.
A system can improve immediate task performance while changing what the user practices. That may be entirely desirable. It may also matter if the bypassed activity was carrying future learning.
Readiness includes the capacity to ask which tasks should be assisted, which should sometimes remain unassisted and how competence will be developed when traditional practice pathways change.
6. Ethical and relational responsibility
AI can produce options. It cannot remove human responsibility merely by making one option easier to select.
In many domains, readiness includes the ability to explain decisions, understand who bears consequences, preserve dignity, recognize conflicts of value and remain answerable when automated assistance has been used.
This is partly individual and partly institutional.
No ethics workshop can compensate for an incentive structure that consistently rewards the opposite behavior.
Readiness changes with consequence
The same person can reasonably use an AI system with very different levels of scrutiny depending on the task.
Generating possible meeting titles is not the same as screening a job applicant.
Summarizing one's own rough notes is not the same as summarizing a child's safeguarding record.
Drafting a decorative image prompt is not the same as supporting a medical decision.
Human readiness is therefore risk-sensitive without being fear-based.
Higher consequence typically increases the importance of verification, traceability, domain expertise, accountability and meaningful human control.
The Institute's framework does not prescribe one universal threshold. It asks that the threshold be visible.
Readiness is institutional before it is personal
A familiar response to technological change is to train the user.
Training matters.
But it can become a convenient way of locating every failure inside the individual.
Suppose employees are taught to verify AI outputs but targets leave thirty seconds per case. Suppose teachers are told to protect independent thinking but assessment systems reward finished products over learning process. Suppose managers are trained in responsible AI but procurement decisions are already locked in. Suppose junior staff are told to challenge automation but the first person who does so is treated as difficult.
The training may be good.
The environment is still teaching something else.
Readiness therefore requires alignment between competence and conditions.
What does cognitive offloading tell us?
Cognitive offloading is normal human behavior.
People write lists, use calculators, set reminders, draw diagrams, consult colleagues and store information outside their heads. Risko and Gilbert's review defines offloading as using physical action or external resources to alter a task's information-processing demands.
Burnett and Richmond's 2026 meta-analysis found benefits of offloading for memory-based task performance under studied conditions, with effects varying by design and task.
This evidence matters because debates about AI often start from the false assumption that retaining human capacity means refusing external support.
It does not.
The relevant problem is selective dependence.
What should be offloaded because the support expands effective capacity?
What still needs rehearsal because future performance depends on internal competence?
What should be duplicated because failure would be costly?
What can be delegated only if a person can still recognize a bad result?
Readiness is the ability to make these distinctions with increasing quality.
A readiness assessment should not become a vanity score
Organizations will be tempted to ask for one number.
“Are we 82 percent AI-ready?”
That number may be useful for a slide and useless for a decision.
A more defensible assessment begins with a real task.
What system is being used?
What decision follows its output?
What must the human understand?
What errors matter?
What authority does the person have?
What capability must persist if the tool is unavailable or wrong?
How is competence learned and tested?
What happens to people affected by the decision?
Which evidence would show that oversight is working rather than merely present?
Readiness becomes measurable when it is tied to function.
What the evidence does not show
Current evidence does not show that:
- there is one validated construct called human readiness for the AI age;
- greater AI use means greater readiness;
- lower AI use means greater wisdom or independence;
- AI literacy alone guarantees good judgment;
- technical capability maps directly onto labor-market displacement;
- people automatically maintain a skill after a tool removes the need to practice it;
- all offloading causes cognitive decline;
- all unaided work produces valuable learning;
- a generic training course can make every role ready for every AI system;
- formal human oversight proves that humans retain meaningful control;
- readiness is solely an individual responsibility.
The Institute's framework should therefore be used as a structured inquiry, not as a claim of scientific consensus.
Institute synthesis: readiness is the human side of deployment
Technology governance asks whether a system is safe enough, compliant enough and technically appropriate to deploy.
Human readiness asks a complementary question.
What must be true of the people and institution around the system for deployment to remain intelligible, contestable and responsible?
The answer may include literacy.
It may include practice.
It may include slowing a workflow.
It may include preserving a task that looks inefficient because the task trains future judgment.
It may include changing who has authority.
It may include deciding not to automate something simply because automation is technically possible.
This is not resistance to AI.
It is readiness for having more of it.
Monday morning
The software still goes live on Monday.
The difference is that the organization has stopped treating go-live as the end of readiness.
People know what the system is for. They know which outputs require independent verification. Junior staff have a route for escalation. Training includes examples in which the correct action is to reject the model's suggestion. Managers monitor not only productivity but error detection, overrides and learning. Some tasks remain deliberately unassisted during apprenticeship. The policy names a human owner for consequential decisions.
None of this makes failure impossible.
It makes responsibility more real.
Human readiness for the AI age is not the promise that people will remain superior to machines at every task.
It is the work of remaining capable of using machines whose abilities may continue to change faster than our institutions do.
References
- OECD. Introducing the OECD AI Capability Indicators. OECD Publishing. 3 June 2025. https://doi.org/10.1787/be745f04-en
- OECD. The OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations. OECD Artificial Intelligence Papers No. 59. 26 May 2026. https://doi.org/10.1787/f3da0f0a-en
- UNESCO. Artificial intelligence and the Futures of Learning: AI Competency Frameworks for Students and Teachers. Frameworks launched during Digital Learning Week 2024. https://www.unesco.org/en/digital-education/ai-future-learning
- European Union. Regulation (EU) 2026/1744 amending Regulation (EU) 2024/1689, revised Article 4 on AI literacy. 8 July 2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj
- Risko EF, Gilbert SJ. Cognitive Offloading. Trends in Cognitive Sciences. 2016;20(9):676-688. https://doi.org/10.1016/j.tics.2016.07.002
- Burnett LK, Richmond LL. Meta-analytic investigations of the effect of cognitive offloading on memory-based task performance and interindividual variability. Memory & Cognition. 2026;54:144-168. https://doi.org/10.3758/s13421-025-01743-8
- Soderstrom NC, Bjork RA. Learning Versus Performance: An Integrative Review. Perspectives on Psychological Science. 2015;10(2):176-199. https://doi.org/10.1177/1745691615569000
- Blume BD, Ford JK, Baldwin TT, Huang JL. Transfer of Training: A Meta-Analytic Review. Journal of Management. 2010;36(4):1065-1105. https://doi.org/10.1177/0149206309352880