RARE SIGNAL | The Nation Inside the Disruption

New Zealand, 28 September 2026

New Zealand is living through a collision of change. Artificial intelligence is moving rapidly from demonstration to daily work. Cybercrime is becoming more forceful, automated and sophisticated. Public services are being asked to deliver more under increasingly constrained resources. The splurge-and-spend well is drying up. Data centres are turning computing capacity into an energy, infrastructure and sovereignty decision. Meanwhile, organisations remain burdened by fragmented systems, accumulated technology debt and inherited processes designed for a slower, more predictable world.

These forces are not arriving neatly or sequentially. They are colliding.

Together, they describe a nation operating in an environment that can now change faster than many organisations can approve their next project. The strategic challenge is therefore no longer simply managing change. It is building organisations capable of learning, deciding and executing at the speed of change itself.

AI has crossed into the operating model

The Government Digital Delivery Agency’s 2026 survey records 545 AI use cases across 59 public-sector organisations, up from 272 across 70 organisations in 2025. Of the 2026 cases, 167 were operational, three times the previous year’s count. Generative AI was the most common technology, followed by natural language processing, agentic AI and machine learning.

The changing respondent count matters: reported use-case growth does not prove that every agency doubled adoption or productivity. Nor does it show whether citizens received better service or risks were controlled.

An AI portfolio needs an owner, decision rights, reliable data, human authority, exception handling, measures of value and a way to stop what fails. Agentic systems sharpen the question: what may the system decide, what must a person approve, and who can reconstruct its actions?

The Government also extended its AI Advisory Pilot on 22 September to 30 June 2027, adding support for 120 small businesses. Around 150 had already joined, and roughly 650 businesses received co-funded places on AI-related courses in the year to June 2026. The next test is whether training changes how work is done.

The cyber threat has already arrived

The National Cyber Security Centre’s 2026 threat report is a hard national signal. It handled 369 incidents of potential national significance in 2025/26. 162 were linked to criminal or financially motivated actors, 18% more than the year before. It recorded four Highly Significant incidents in one year, as many as in the previous decade combined.

The SME behaviour tracker exposes the gap between threat and preparedness. 43% of SMEs now believe they are vulnerable, up from 34%. Among firms with 20–49 employees, 76% reported a cyber threat or attack in the preceding six months; among medium-sized firms that experienced an attack, 44% reported moderate-to-severe impact. Yet 32% of SMEs were taking no action to upskill staff in cyber security.

The NCSC warns that frontier AI can increase the speed, scale and precision of attacks. Its judgement that advanced capabilities may reach malicious actors by early 2027 is a planning scenario, not a prediction of a particular incident. The response: identify critical systems and suppliers; enforce identity controls; patch; detect unusual behaviour; test recovery; rehearse decisions. A policy cannot restore a service. Practised execution can.

Rare’s internal Global Cyber Disruption Index stands at 77/100 this week, in its Severe band. That is an analytical risk indicator, not an official statistic or a probability of attack. Its value lies in forcing explicit attention to actor tempo, shared technology, critical services and recovery strain.

Digital investment faces an ownership test

The Public Service Commission’s digital reset review found fragmented investment, duplication and limited coordination across government. That diagnosis should sound familiar well beyond Wellington. Too many organisations buy platforms while leaving process design, information ownership, integration, staff capability and benefits measurement unresolved.

An execution example also requires scrutiny. One NZ reported on 16 September that its enterprise mobile provisioning fell from up to ten days to under ten minutes after orchestrating systems and human work. It is a company-reported result; volumes, exception rates, quality and unit economics still matter. The workflow changed across boundaries. A faster screen alone would not have delivered it.

The question for every CEO and board: after consultants and vendors leave, who owns the architecture, data, controls, process knowledge and proof of value? If those remain outside the enterprise, the invoices may be the only durable artefact.

Health shows progress—and the limits of a headline

On 23 September, the Government reported improvement across all five national health targets. The proportion of patients waiting under four months for elective treatment rose from 63.9% to 72.6% year on year; first specialist assessments from 62.0% to 66.1%; and six-hour emergency-department performance from 73.9% to 76.3%. Cancer treatment within 31 days reached 87.0%, and immunisation by age two 83.7%. More than 14,200 fewer people were waiting over four months for elective treatment than at the peak.

That progress matters. It does not establish which intervention produced which gain, whether benefits reached every locality or population fairly, or what happened to workforce load, clinical quality and patient experience. For health, hospice and community care, the useful intelligence runs from demand and referral through staffing, hand-off, treatment, continuity and outcome. Digital work must serve that chain. A dashboard is an instrument, not a clinical result.

The world is reshaping the cost base

The International Energy Agency’s April 2026 outlook projects data-centre electricity use rising from 485 TWh in 2025 to about 950 TWh by 2030, around 3% of global demand. The local capital signal is visible: Infratil disclosed on 15 September that data centres accounted for just over half its NZ$22 billion asset value; CDC had 350 MW deployed and 1.1 GW contracted. Those are disclosed positions and projections, not proof that every planned megawatt will arrive on time.

Meanwhile, the World Economic Forum’s employer-based forecast estimates that creation and displacement could affect 22% of today’s formal jobs by 2030 - 170 million created, 92 million displaced. This is a scenario built from employer expectations, not a count of jobs already lost. Its signal is the scale of capability change organisations must manage.

New Zealand cannot treat energy, skills, cyber, clinical capacity and AI as separate conversations. They compete for capital, leadership attention and public trust. Strong organisations will question the problem, redesign work, establish accountable data and controls, measure change, and revise strategy as the environment shifts.

Finally, sectors will survive, thrive or fail according to the contest within them - that inner turmoil or institutional calcification against the recognition that change is required. Inherited incentives, rigid governance and comfortable (or uncomfortable) operating models can make an obvious problem impossible to act on. Where those inert forces win, disruption becomes a crisis imposed from outside - except when the 'for lease' signs go up and doors have to shut. Where leaders break their grip, change becomes a capability the sector can use.

Execution is the dividing line. Chatter has become an industry - panels, policies, decks, launch statements and endless declarations of intent. It can fill a calendar and empty a budget without shortening a treatment wait, improving a student’s learning, housing a family, clearing a consent backlog, closing a cyber exposure, connecting new power or delivering a dollar of productivity or outcome.

An announced pilot is not adoption. Policies without execution or use-cases, yeah nah.

A dabble, here, there and not much anywhere....reported token activity is not an outcome. The test is brutal - what changed for citizens, patients, workers, customers and the balance sheet—and who can prove it? In a disruptive era, the rare advantage is turning evidence into decisions, decisions into execution, and execution into multi-layered outcomes the organisation owns.

Everything else is noise with an inert decaying budget.