Wellington AI in Government Summit 2027

Beat the rush and join us early for complimentary barista-made coffee and breakfast.

With a public service smaller than many comparable jurisdictions and genuine constraints on the scale of investment agencies can make individually, this opening keynote examines how New Zealand government can build durable AI capability and shared infrastructure without simply importing approaches designed for much larger, better-resourced systems.

• How individual agency AI investment, replicated across dozens of smaller agencies, risks producing costly duplication that a country of New Zealand's size can't sustain.
• Lessons from existing shared digital infrastructure efforts led by the Government Chief Digital Officer and Digital.govt.nz on what it actually takes to build AI capability once and use it many times across government.
• How a small public service maintains genuine oversight and negotiating leverage over AI vendors whose scale and market power dwarf that of any individual New Zealand agency.

As government agencies deploy AI systems that touch on data, services, and decisions affecting Māori communities, this keynote examines what Te Tiriti o Waitangi's partnership principles genuinely require in AI governance, moving beyond token consultation toward Māori data sovereignty and co-design as foundational practice.

• Unpacking the specific obligations agencies face around how Māori data is collected, used, and governed within AI systems, distinct from general privacy compliance.
• Examples of where agencies have involved Māori communities and data governance experts in shaping AI systems from the outset, rather than seeking feedback on decisions already made.
• What it takes for public servants to genuinely understand and apply te ao Māori perspectives in AI design, beyond satisfying a partnership checkbox.

As government services increasingly channel citizens toward AI-powered digital platforms, this panel examines the specific access and equity challenges facing rural New Zealanders, Pacific communities, and older citizens, populations whose service needs and connectivity realities differ significantly from the urban, digitally-connected default many AI systems are designed around.

• How patchy rural broadband and mobile coverage undermine the basic assumption behind AI-first service design, and what agencies are doing to design around that reality rather than ignore it.
• How agencies are engaging Pacific communities directly to understand where AI-driven services are failing to meet cultural and linguistic needs, beyond simple translation.
• How agencies are tracking whether rural, Pacific, and older citizens are actually achieving equivalent service outcomes through AI channels, rather than assuming access alone solves the equity problem.

After realising that Algorithm Charter compliance had become a documentation exercise completed after systems were already built, one agency's data and analytics leader rebuilt their development process to embed the Charter's principles into design decisions from the very first sprint.

• How the team discovered that Charter assessments were being completed to satisfy process requirements after key design decisions were already locked in, rather than shaping them.
• The specific changes made to bring transparency, bias testing, and human oversight considerations into early design conversations, rather than a late-stage compliance review.
• How catching design issues early, rather than after a completed build, ultimately reduced costly rework and helped the team deliver faster, not slower.

A hands-on, interactive session working through a real public sector AI scenario as a room. Details announced soon.

After an Official Information Act request asking how an AI system had contributed to a specific decision drew a technically accurate but practically useless response, one agency's information and technology leaders rebuilt their approach so future OIA requests about their AI systems could be answered substantively, not deflected with jargon.

• How the team recognised that a response citing model architecture and confidence scores technically answered the request while leaving the requester no clearer on why the decision was actually made.
• The specific changes made to how the system recorded its reasoning, so that information could later be translated into a plain-language explanation rather than reconstructed after the fact under time pressure.
• How the team created a repeatable approach for translating technical AI system behaviour into OIA-appropriate language, rather than treating each request as a bespoke problem.

Recognising that three agencies were each independently trying to build similar AI capability with limited budgets and overlapping needs, one government digital leader built a shared capability model that let the agencies pool investment and expertise rather than duplicating effort three times over.

• How the team recognised that each agency was independently procuring similar AI tools and building similar internal skills, at a combined cost none of them could justify alone.
• How the model balanced genuine shared capability with the flexibility each agency needed for its own specific use cases and legislative context.
• The negotiation and trust-building required to get agencies accustomed to independent decision-making to commit to a shared model.

As New Zealand government agencies embed AI into core service delivery, this panel examines the emerging cybersecurity risks specific to AI systems, from data poisoning to model manipulation, and whether the country's cyber resilience settings, shaped for a smaller and less AI-dependent government, are keeping pace.

• Understanding threats specific to AI deployments that traditional government cybersecurity frameworks weren't designed to address.
• Assessing whether current National Cyber Security Centre guidance adequately addresses AI-specific risks, or whether agencies are filling gaps themselves.
• How agencies with limited security capability are prioritising AI-specific risk against a broader cyber threat environment competing for the same scarce resources.

Small-group, discussion-based sessions where you'll work through real public sector AI challenges with peers in similar roles. Roundtable topics will be announced soon.

Put your knowledge to the test in this fast-paced quiz covering real-world trivia, key concepts, and emerging trends. Compete for bragging rights — and a voucher — as the top scorer takes the crown.

Facing a global AI vendor's standard contract terms that offered New Zealand agencies little genuine control over data handling or model behaviour, one procurement and legal leader built a negotiating approach that secured meaningfully better sovereignty protections despite New Zealand's relatively small market leverage.

• How the team identified the specific gaps in the vendor's default data handling and audit rights that posed genuine risk for government use cases.
• How aggregating demand across multiple agencies, rather than negotiating individually, gave the team meaningfully more influence than any single agency could achieve alone.
• How the team decided which sovereignty protections were non-negotiable for government use, even if that meant a longer negotiation or a smaller shortlist of viable vendors.

With New Zealand's social sector agencies among the most experienced in the world at using predictive analytics and risk modelling for frontline service decisions, this keynote draws out hard-won lessons about the boundary between genuinely helpful predictive tools and systems that risk entrenching disadvantage.

• Honest reflection on where predictive analytics in social services has improved outcomes for vulnerable New Zealanders, and where it has drawn justified criticism for reinforcing bias.
• How leading agencies are designing systems that inform frontline judgement rather than substitute for it, particularly in decisions affecting children and vulnerable families.
• What other agencies now deploying AI in frontline services, from immigration to justice to health, can learn from the social sector's longer history with algorithmic decision support.

As AI systems increasingly generate recommendations for public servants to act on, this interactive discussion examines a subtle but consequential risk: whether staff are exercising genuine judgement over AI outputs, or quietly deferring to the machine even when their own expertise suggests otherwise.

• Evidence and experience suggesting that staff under time pressure or facing confidently-presented AI recommendations often approve rather than genuinely scrutinise them.
• What agencies are learning about interface design, workload, and culture changes that make staff more likely to question AI outputs rather than defer to them by default.
• How some agencies are tracking override and disagreement rates between staff and AI recommendations as an early warning signal for over-reliance, rather than assuming a smoothly running system is a healthy one.

Unwind with your peers for a couple of drinks on us!

Beat the rush and join us early for complimentary barista-made coffee and breakfast.

With a public service smaller than many comparable jurisdictions and genuine constraints on the scale of investment agencies can make individually, this opening keynote examines how New Zealand government can build durable AI capability and shared infrastructure without simply importing approaches designed for much larger, better-resourced systems.

• How individual agency AI investment, replicated across dozens of smaller agencies, risks producing costly duplication that a country of New Zealand's size can't sustain.
• Lessons from existing shared digital infrastructure efforts led by the Government Chief Digital Officer and Digital.govt.nz on what it actually takes to build AI capability once and use it many times across government.
• How a small public service maintains genuine oversight and negotiating leverage over AI vendors whose scale and market power dwarf that of any individual New Zealand agency.

As government agencies deploy AI systems that touch on data, services, and decisions affecting Māori communities, this keynote examines what Te Tiriti o Waitangi's partnership principles genuinely require in AI governance, moving beyond token consultation toward Māori data sovereignty and co-design as foundational practice.

• Unpacking the specific obligations agencies face around how Māori data is collected, used, and governed within AI systems, distinct from general privacy compliance.
• Examples of where agencies have involved Māori communities and data governance experts in shaping AI systems from the outset, rather than seeking feedback on decisions already made.
• What it takes for public servants to genuinely understand and apply te ao Māori perspectives in AI design, beyond satisfying a partnership checkbox.

As government services increasingly channel citizens toward AI-powered digital platforms, this panel examines the specific access and equity challenges facing rural New Zealanders, Pacific communities, and older citizens, populations whose service needs and connectivity realities differ significantly from the urban, digitally-connected default many AI systems are designed around.

• How patchy rural broadband and mobile coverage undermine the basic assumption behind AI-first service design, and what agencies are doing to design around that reality rather than ignore it.
• How agencies are engaging Pacific communities directly to understand where AI-driven services are failing to meet cultural and linguistic needs, beyond simple translation.
• How agencies are tracking whether rural, Pacific, and older citizens are actually achieving equivalent service outcomes through AI channels, rather than assuming access alone solves the equity problem.

After realising that Algorithm Charter compliance had become a documentation exercise completed after systems were already built, one agency's data and analytics leader rebuilt their development process to embed the Charter's principles into design decisions from the very first sprint.

• How the team discovered that Charter assessments were being completed to satisfy process requirements after key design decisions were already locked in, rather than shaping them.
• The specific changes made to bring transparency, bias testing, and human oversight considerations into early design conversations, rather than a late-stage compliance review.
• How catching design issues early, rather than after a completed build, ultimately reduced costly rework and helped the team deliver faster, not slower.

A hands-on, interactive session working through a real public sector AI scenario as a room. Details announced soon.

After an Official Information Act request asking how an AI system had contributed to a specific decision drew a technically accurate but practically useless response, one agency's information and technology leaders rebuilt their approach so future OIA requests about their AI systems could be answered substantively, not deflected with jargon.

• How the team recognised that a response citing model architecture and confidence scores technically answered the request while leaving the requester no clearer on why the decision was actually made.
• The specific changes made to how the system recorded its reasoning, so that information could later be translated into a plain-language explanation rather than reconstructed after the fact under time pressure.
• How the team created a repeatable approach for translating technical AI system behaviour into OIA-appropriate language, rather than treating each request as a bespoke problem.

Recognising that three agencies were each independently trying to build similar AI capability with limited budgets and overlapping needs, one government digital leader built a shared capability model that let the agencies pool investment and expertise rather than duplicating effort three times over.

• How the team recognised that each agency was independently procuring similar AI tools and building similar internal skills, at a combined cost none of them could justify alone.
• How the model balanced genuine shared capability with the flexibility each agency needed for its own specific use cases and legislative context.
• The negotiation and trust-building required to get agencies accustomed to independent decision-making to commit to a shared model.

As New Zealand government agencies embed AI into core service delivery, this panel examines the emerging cybersecurity risks specific to AI systems, from data poisoning to model manipulation, and whether the country's cyber resilience settings, shaped for a smaller and less AI-dependent government, are keeping pace.

• Understanding threats specific to AI deployments that traditional government cybersecurity frameworks weren't designed to address.
• Assessing whether current National Cyber Security Centre guidance adequately addresses AI-specific risks, or whether agencies are filling gaps themselves.
• How agencies with limited security capability are prioritising AI-specific risk against a broader cyber threat environment competing for the same scarce resources.

Small-group, discussion-based sessions where you'll work through real public sector AI challenges with peers in similar roles. Roundtable topics will be announced soon.

Put your knowledge to the test in this fast-paced quiz covering real-world trivia, key concepts, and emerging trends. Compete for bragging rights — and a voucher — as the top scorer takes the crown.

Facing a global AI vendor's standard contract terms that offered New Zealand agencies little genuine control over data handling or model behaviour, one procurement and legal leader built a negotiating approach that secured meaningfully better sovereignty protections despite New Zealand's relatively small market leverage.

• How the team identified the specific gaps in the vendor's default data handling and audit rights that posed genuine risk for government use cases.
• How aggregating demand across multiple agencies, rather than negotiating individually, gave the team meaningfully more influence than any single agency could achieve alone.
• How the team decided which sovereignty protections were non-negotiable for government use, even if that meant a longer negotiation or a smaller shortlist of viable vendors.

With New Zealand's social sector agencies among the most experienced in the world at using predictive analytics and risk modelling for frontline service decisions, this keynote draws out hard-won lessons about the boundary between genuinely helpful predictive tools and systems that risk entrenching disadvantage.

• Honest reflection on where predictive analytics in social services has improved outcomes for vulnerable New Zealanders, and where it has drawn justified criticism for reinforcing bias.
• How leading agencies are designing systems that inform frontline judgement rather than substitute for it, particularly in decisions affecting children and vulnerable families.
• What other agencies now deploying AI in frontline services, from immigration to justice to health, can learn from the social sector's longer history with algorithmic decision support.

As AI systems increasingly generate recommendations for public servants to act on, this interactive discussion examines a subtle but consequential risk: whether staff are exercising genuine judgement over AI outputs, or quietly deferring to the machine even when their own expertise suggests otherwise.

• Evidence and experience suggesting that staff under time pressure or facing confidently-presented AI recommendations often approve rather than genuinely scrutinise them.
• What agencies are learning about interface design, workload, and culture changes that make staff more likely to question AI outputs rather than defer to them by default.
• How some agencies are tracking override and disagreement rates between staff and AI recommendations as an early warning signal for over-reliance, rather than assuming a smoothly running system is a healthy one.

Unwind with your peers for a couple of drinks on us!