Agent-Ready New Zealand

Preparing businesses, services, and public institutions for the next interface of the global economy

The next interface is being wired into the global economy

Much of the organisational focus on AI still sits inside the business.

That is the current centre of gravity: productivity, analysis, service support, software development, workflow automation, and faster access to information. Global survey evidence shows broad organisational adoption, but most organisations remain in experimentation or pilot phases rather than scaled enterprise-wide deployment.[1]

These are important areas of work, and they are where organisations can see the clearest value.

Another layer is beginning to form at the boundary between people and organisations. AI is becoming part of how people search, compare, interpret, choose, act, challenge, and manage relationships with services.

A customer may no longer begin with a website.

A citizen may not begin with a form.

A student may not begin with a course guide.

A traveller may not begin with an airline, a search engine, or a booking platform.

They may begin with an AI-generated answer or an agent acting on their behalf.

That changes the interface.

In WAVES, I describe this as the movement from AI inside organisations to AI between people and organisations. The first wave is about internal capability: people using AI to support work. The second begins when AI starts to sit between people and the organisations they rely on.

This essay focuses on that second-wave interface, where people, organisations, agents, platforms, data, identity, authority, and correction pathways meet. Trusted by Design develops the wider trust architecture.

Agent-readiness belongs in this second wave. It asks whether an organisation can still be found, interpreted, verified, accessed, corrected, and trusted when AI becomes part of the interaction.

Development will be uneven, and existing channels will remain important. Organisations still need to prepare for another pathway alongside them.

When AI represents the person

The first side of this interface is the person.

That person might be a customer, citizen, student, patient, traveller, employee, parent, homeowner, investor, business owner, or exporter.

They may not think of themselves as using an “agent”. They may simply ask an AI system to help them get something done.

OpenAI’s developer documentation describes agents as applications that can plan, call tools, collaborate across specialists, and keep enough state to complete multi-step work.[2] That is the important distinction here: agents are not just conversational interfaces. They can become task interfaces.

Consumer agent products already combine research, web interaction, scheduling and booking, and other authorised actions, although their availability, reliability and use remain uneven.[3]

Compare these providers.

Tell me what a normal price range should be.

Find the best option for my situation.

Tell me what I’m eligible for.

Book the appointment.

Challenge the charge.

Prepare the form.

Summarise what I need to know before I decide.

Pragmatically, this is how the change will show up.

A traveller could compare airlines, fares, baggage rules, loyalty benefits, disruption policies, and refunds before visiting an airline’s website.

A homeowner could move through council consent, builders, insurers, finance, and compliance with an agent interpreting each step.

A student could compare courses, fees, credentials, work rights, scholarships, and career pathways through one guided interaction.

A small business could use an agent to interpret tax obligations, banking, logistics, insurance, software, grants, and export support.

In each case, the person is still making decisions.

The path into the service is different.

The agent becomes part of how the person understands the options, forms an opinion, takes action, and asks for help when something goes wrong. It may also influence which options appear, how they are framed, and which defaults are easiest to accept.

Human agency includes understanding the agent’s role, authorising and confirming actions, challenging the result, withdrawing permission, and reaching human support.

I think this will be one of the more important service changes for organisations to understand.

AI will alter how people arrive, what they already know, what they expect to be able to do, and how quickly they expect uncertainty to be resolved. The same preparation can improve today’s digital services even where agent-mediated use develops more slowly.

When AI represents the organisation

The second side of the interface is the organisation.

If people use AI agents to navigate services, organisations need to be understandable to those agents.

That sounds simple.

It probably won’t be.

A website written for people is useful. A brand promise is useful. A PDF policy is useful. A traditional contact centre is useful.

Those channels were designed for a world where people came through familiar pathways and did much of the interpretation themselves.

Agent-mediated interaction puts new pressure on the organisation.

Information needs to be clear enough for intelligent systems to interpret.

Services need to be structured enough for intelligent systems to navigate.

Authority needs to be clear enough for intelligent systems to respect.

Errors need to be correctable when intelligent systems get something wrong.

The practical starting point is readiness. Building a sophisticated organisational agent can come later.

AI readiness and agent-readiness are related, but they address different questions.

Agent-readiness focuses on making services understandable, usable, and accountable when AI becomes part of the interaction. Human service remains part of that design.

Many foundations are familiar: current content, structured data, identity, service design, escalation, and review. Agent-readiness joins them around a new condition: an intelligent intermediary may interpret, combine, and act on organisational information before the person arrives.

For a business, this may mean clearer product data, pricing, eligibility, service terms, support pathways, and correction processes. It may also mean publishing information in more structured ways, reducing ambiguity in terms and conditions, keeping customer-facing content current, and making it easier for people or agents to verify what is true.

For a public institution, it may mean forms, rules, obligations, entitlements, complaints, review pathways, and human escalation points that can be interpreted more reliably. People should be able to understand what applies, what evidence is required, what can be reviewed, and where responsibility sits.

For an exporter, it may mean making products, provenance, certifications, sustainability claims, logistics information, and market access requirements easier for AI systems to understand without distortion. Trusted product data, verified claims, and authoritative market information become part of export readiness.

For an iwi organisation, it may mean expressing data governance, consent, authority, cultural context, and benefit expectations in ways digital systems can respect without flattening the relationships behind them. It may also mean clearer boundaries around representation, permission, restriction, and who has authority to speak for particular data, knowledge, or relationships. Te Mana Raraunga establishes that Māori data should be subject to Māori governance and extends those rights and interests into algorithmic systems.[4][5]

Agent-readiness is selective legibility, not unrestricted openness. Security, privacy, commercial sensitivity, cultural authority, and governed knowledge remain prior conditions. The right information and pathways should be understandable under the right authority.

The language of “agent-ready” gives leaders a way to look beyond chatbots and ask more precise questions:

Can our organisation be found and understood properly by AI systems?

Can those systems recognise trusted information, respect authority, and move through the right correction pathways when something is wrong?

Can our organisation participate in the agents, platforms, and digital interfaces that people will use to discover, compare, access, and manage services?

Trust has to run through the interface

When AI becomes part of the interface, trust changes shape. It has to work inside the interface itself.

A customer may ask an agent to compare providers. The agent needs to know which information is current.

A person may ask a system to recommend the best option. The system needs to know which claims can be trusted.

A business may want to appear in a comparison. The interface needs to understand what the business offers, what it stands behind, and where responsibility sits.

A person may challenge a charge, decision, booking, or service outcome. The agent needs a correction pathway it can follow.

This is where the interface begins to affect growth, service quality, customer access, and trust.

If AI systems become part of how people discover, compare, select, buy, renew, complain, and manage services, trust becomes part of customer access.

Trust shapes visibility.

It shapes interpretation.

It shapes whether claims are treated as reliable, whether a customer can move from discovery to action, and whether errors can be corrected before they damage the relationship.

In the existing interface, trust often sits around the interaction. A person can recognise a brand, read a policy, call a contact centre, ask a professional, or complain after something goes wrong.

Those pathways still matter. Agent-mediated interaction adds another layer.

AI systems can interpret information before a person sees it. They can combine sources, summarise options, act on instructions, and make some organisations easier or harder to reach.

The organisation may lose the person’s trust before a human conversation begins.

A recent Google AI Overviews case provides a concrete signal. Reuters reported on 12 June 2026 that Google planned to appeal a Munich court ruling that held it legally liable for false claims appearing in AI Overviews and treated the generated summaries as Google’s own content.[6]

The case is a jurisdiction-specific signal under appeal, rather than a complete answer to where responsibility sits. The wider service issue is visible: generated interfaces can place organisational information, platform interpretation, and a person’s decision inside the same interaction.

An organisation cannot control how every external model retrieves, ranks, or summarises its information. Readiness offers influence rather than control. It gives the organisation stronger authoritative sources, clearer expressions of permission and responsibility, and a visible route for correction when the interface gets something wrong.

This is where Agent-Ready New Zealand connects with my recent white paper Trusted by Design.

Trust has to travel through the information an agent reads, the authority it recognises, the permission it relies on, the action it takes, and the correction pathway available when something is wrong.

For agent-ready organisations, trust becomes part of growth, value, and customer interaction. It shapes how the organisation participates in the next interface of the digital economy.

New Zealand in a globally shaped interface

For New Zealand, the question becomes larger.

How will our businesses, public services, exporters, iwi organisations, universities, regions, products, credentials, rules, and places be understood when AI-generated interfaces sit between them and the people trying to reach them?

A country can have good services and still be poorly represented.

A business can have a strong offer and still be misread by AI.

A public institution can have the right policy and still be difficult for intelligent systems to interpret.

This is already visible in smaller ways. People use search engines, maps, reviews, comparison sites, booking platforms, social media, and digital marketplaces to decide what to trust. AI-generated interfaces add another layer to that pattern. They can summarise, recommend, compare, and act. They can also influence which pathways are visible, which claims appear credible, and which organisations are easiest to reach.

Global platforms, model providers, search systems, marketplaces, agent frameworks, data standards, commercial incentives, and the information available to those systems will shape much of this interface.

New Zealand organisations cannot determine all of those conditions. They can influence the quality, provenance, authority, accessibility, and correction pathways of the information those systems encounter.

That is the organisational-to-national connection.

When the same readiness conditions recur across public services, exporters, businesses, universities, iwi organisations, credentials, and places, some dependencies become shared. Digital identity, trusted credentials, authoritative public information, provenance, common service standards, language capability, data governance, and visible routes for correction extend beyond a single organisation.

New Zealand’s 2025 AI strategy emphasises adoption and application rather than foundational AI development, alongside international engagement and science and innovation capability.[7] Agent-readiness adds a specific interface question to that wider direction.

Agent-readiness therefore has a national dimension. In NZ-EOS terms, it forms part of the interface layer between people, organisations, platforms, trust systems, and economic value.

If New Zealand wants to create and retain value in the intelligence economy, its organisations need to be visible and usable in the interfaces where future decisions are made.

This affects customers choosing services, investors assessing credibility, tourists comparing destinations, students comparing education pathways, exporters in global markets, and people accessing public support, rights, services, or review.

International evaluation shows persistent performance gaps between English and languages with less representation in training data and benchmarks, including weaknesses in mixed-language and culturally sensitive tasks.[8] Publicly available evidence does not yet provide a robust basis for claiming how current systems perform across te reo Māori, specific Pacific languages, regional New Zealand usage, or culturally specific requests. Those capabilities need local evaluation, appropriate authority, and practical routes to correct misrepresentation.

Greater national readiness can develop without one central programme or local control over global platforms. It requires recognising where individual preparation depends on shared infrastructure, standards, evaluation, governance, or coordination.

New Zealand can still choose how prepared its organisations are to participate.

Without clearer information, trusted identity and credentials, governed authority, local evaluation, and visible correction pathways, New Zealand risks being represented through systems it does not shape, using information it does not control, in interfaces where its organisations may be misread, bypassed, or made harder to reach.

With those foundations strengthened across organisations and shared systems, New Zealand has a better chance of being visible, trusted, and usable in the places where future decisions are made.

The foundations of agent-readiness

Agent-readiness prepares New Zealand organisations for a world where intelligent systems help people discover, interpret, compare, act, challenge, and manage relationships with services.

Preparation begins with clearer information.

Products, services, prices, eligibility, obligations, support pathways, locations, policies, exceptions, and review processes need to be easier for people and systems to understand. Authoritative sources should show what is current, what has changed, and where responsibility sits.

Agent-readiness also depends on verifiable identity and credentials.

New Zealand already has a foundation to build on. The Digital Identity Services Trust Framework Act 2023 establishes a legal framework for secure and trusted digital identity services for individuals and organisations, including governance and accreditation functions. The wider trust framework sets rules for accredited digital identity services and the protection of information and privacy.[9][10]

People need to prove who they are. Organisations need to prove who they are. Representatives, licences, qualifications, certifications, permissions, and claims need trusted ways to be checked without forcing people to share more than necessary.

Consent and delegated authority become more important as well.

If an agent is acting for a person, organisation, or platform, the boundaries need to be clear. What can it do? Who gave permission? What evidence supports that authority? When does the action need human confirmation? How can permission be narrowed or withdrawn?

These foundations work together. Clear information without authority can mislead. Identity without consent can enable the wrong action. Machine-readable pathways without human review can make a service difficult to challenge. Readiness comes from the relationship between information, identity, permission, action, and correction.

Correction, recourse, and accountability

In the current digital interface, a person might notice an error on a website, call a contact centre, email support, lodge a complaint, or update a form.

In an agent-mediated interface, the error may appear earlier and spread differently.

An AI system might summarise the wrong policy.

It might compare an outdated price.

It might misunderstand eligibility.

It might recommend the wrong pathway.

It might carry forward a mistake from one system into another.

Correction therefore has to be easier for both people and systems to find.

Organisations will need authoritative sources of information that AI systems can rely on. They will need visible ways to signal what is current, what is verified, what has changed, and where an error should be corrected.

Human review and recourse also need to work through the new interface.

The Public Service AI Work Programme provides an early practical example. The two-year programme includes a Govt.nz AI Assistant deliverable, described as an AI search tool intended to help people find government services and information through a conversational interface. The programme also includes initiatives across common tools, safe and responsible AI, customer and partnerships, and workforce capability.[11]

If a person is affected by an AI-mediated interaction, they should not be trapped inside a chain of automated responses. They need a way to move from the agent, platform, or generated answer back to a responsible organisation.

Who can check the outcome?

Who can explain the decision?

Who can correct the record?

Who can override the automated pathway?

Who carries responsibility when the interface gets it wrong?

This is where governance becomes visible to the customer. It is not just a board policy, a risk framework, or an internal assurance process. It shows up in whether a person can reach the right pathway when something goes wrong.

Leaders also need clear accountability.

Agent-mediated interaction will blur boundaries between customers, organisations, platforms, software providers, and intermediaries. That makes ownership more important, not less. The legal allocation of responsibility will vary by jurisdiction, contract, role, and context. The operational requirement remains: a person needs an identifiable route to the organisation responsible for the service, information, or decision.

If a customer’s agent interacts with a company’s system through a platform interface, the organisation still needs to know which parts of that experience it owns, which parts it can influence, and where the customer can get help.

For many organisations, this changes the question.

“How do we use AI?” remains useful. Agent-readiness adds another important consideration: whether the organisation can be found, understood, trusted, and reached through AI-mediated interfaces, with authority and correction intact.

That is the practical meaning of agent-readiness. It gives businesses, services, institutions, and public systems a way to prepare for the next interface of the digital economy.

If AI becomes part of the wiring between people and organisations, trust, authority, correction, and accountability have to move through that wiring as well.

References

1. McKinsey & Company, “The state of AI in 2025: Agents, innovation, and transformation,” 5 November 2025.
2. OpenAI, “Agents SDK,” accessed 22 July 2026.
3. OpenAI, “Introducing ChatGPT agent: bridging research and action,” 17 July 2025.
4. Te Mana Raraunga, “What is Māori Data Sovereignty?,” accessed 22 July 2026.
5. Te Mana Raraunga, “Indigenous Data Sovereignty, AI & Algorithms,” accessed 22 July 2026.
6. Foo Yun Chee, “Google to challenge German ruling saying it is liable for AI-generated false claims,” Reuters, 12 June 2026.
7. Ministry of Business, Innovation & Employment, “New Zealand’s AI Strategy: Investing with confidence,” July 2025.
8. Wenhan Han, Yifan Zhang, Zhixun Chen, Binbinliu, Mykola Pechenizkiy, Meng Fang, and Yin Zheng, “MuBench: Assessment of Multilingual Capabilities of Large Language Models Across 61 Languages,” Findings of the Association for Computational Linguistics: ACL 2026, July 2026, doi:10.18653/v1/2026.findings-acl.794.
9. New Zealand Legislation, “Digital Identity Services Trust Framework Act 2023,” current version accessed 22 July 2026.
10. New Zealand Digital Government, “Trust framework for digital identity,” last updated 19 September 2024, accessed 22 July 2026.
11. New Zealand Digital Government, “Public Service AI Work Programme,” last updated 16 January 2026, accessed 22 July 2026.

Related Frameworks and White Papers

• WAVES: A framework describing AI’s movement from work inside organisations, to interaction between people and organisations, and then into wider agent-mediated networks.
• Trusted by Design: A white paper on trust, identity, consent, delegated authority, provenance, assurance, accountability, human review, and redress in the intelligence economy.
• New Zealand Economic Operating System (NZ-EOS): A national systems framework connecting infrastructure, trust, capability, institutions, exports, and long-duration economic positioning.
• The Machine Room: A framework for the physical, digital, trusted, institutional and human foundations required for intelligence-era services and economic capability.

Related Essays

• Trust as New Zealand’s Economic Capability
• Building AI Advantage on Sovereign Data and Trust
• A Practical Way to Scale AI Across Your Business: The Studio Model