Sydney AI in FSI Summit 2027

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

As financial institutions move from AI that recommends to AI that autonomously executes; approving loans, rebalancing portfolios, settling claims, this keynote examines what genuinely changes when decisions once made by humans start being made by agents, and sets the tone for a day exploring what that shift demands of governance, culture, and trust.

• Tracing the real trajectory of AI's role in financial institutions over the past few years, and why 2027 marks a genuine inflection point rather than incremental progress.
• The governance, audit trail, and human-oversight infrastructure that needs to exist before an institution should let an AI agent act without a human in the loop.
• Where does agentic AI genuinely create value versus introduce unacceptable risk, and how do institutions build the muscle to tell the difference?

Scam prevention is moving from detection to real-time intervention, with banks increasingly expected to decide whether to warn, delay or block a payment while the customer is still trying to make it. The challenge is getting that intervention right without creating unnecessary friction for legitimate customers.

This keynote explores how financial institutions are redefining payment intervention in the era of increasingly sophisticated scams, including how they set thresholds for action, respond when customers insist a payment is genuine, and navigate responsibility when warnings are ignored.

• What signals and behaviours should trigger a warning, delay or payment block?
• How should banks respond when customers have been coached by scammers and remain convinced the payment is legitimate?
• What happens when a bank identifies the risk, warns the customer, but ultimately allows the payment to proceed?

An AI agent connected to internal systems may have access to customer records, APIs, documents, code and operational tools. Giving an agent too much access can create a much larger problem than an inaccurate chatbot answer.

This panel looks at how financial institutions are thinking about identities, permissions and access controls for AI agents that need to take real actions.We'll discuss:

• Should every agent have its own identity, credentials and clearly defined permissions?
• Which systems should agents never access without a fresh human approval?
• How do you detect when an agent is using legitimate access in a way nobody intended?

As AI models proliferated across the organisation faster than the model risk team could individually review them, one risk leader rebuilt their governance framework around tiered, risk-proportionate oversight, maintaining rigour for high-stakes models without creating a bottleneck for lower-risk use cases.

• How the team identified that applying the same validation rigour to every model, regardless of stakes, was slowing adoption without improving safety.
• The practical criteria used to categorise models by potential impact, and how those tiers determined the depth of review and monitoring required.
• How automated monitoring and documentation tools let a stable risk team keep pace with a rapidly growing model inventory.

A hands-on, interactive session working through a real AI in financial services scenario as a room. Details announced soon.

Facing the challenge of deploying a sophisticated machine learning credit model while maintaining genuine explainability across very different audiences, one lending leader built an explainability layer that satisfied regulatory scrutiny, customer transparency requirements, and legal defensibility simultaneously.

• How the team designed distinct explanation outputs for regulators, customers, and internal dispute resolution, rather than a single generic explanation.
• The trade-offs made between model complexity and explainability, and where the team drew the line.
• How the team stress-tested their explanation framework against real dispute scenarios before deploying the model into production.

After an initial AI-driven claims automation project succeeded in speed but drew criticism for depersonalised handling of complex cases, one insurance leader rebuilt the system to combine automation's efficiency with better outcome measurement and appropriate human escalation.

• How customer feedback and complaint data revealed that speed gains were masking a decline in outcome quality for complex claims.
• How the team redesigned the routing logic to identify which claims genuinely needed human judgement, regardless of dollar amount.
• The shift from processing-time metrics to outcome and satisfaction measures that better reflected genuine customer experience.

AI can help members and customers understand their finances, model options and receive increasingly personalised guidance. The difficult point is deciding when useful assistance crosses into regulated financial advice.This panel looks at where banks, super funds and wealth organisations are drawing that line and how they are designing human escalation into digital advice journeys.We'll discuss:

• What can an AI assistant personalise without crossing into regulated advice?
• Which questions or customer circumstances should automatically trigger a licensed adviser?
• How do organisations scale digital guidance while preserving enough context for a human to take over safely?

Small-group, discussion-based sessions where you'll work through real AI in financial services 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.

AI coding tools are increasing the amount and speed of software produced inside financial institutions. Testing, security review and change controls still have to keep pace with code that may now be partly generated by an agent rather than written line by line by an engineer.This session looks at how one financial institution is using AI across engineering while adapting its review, testing and release controls for a much faster development cycle.We'll discuss:

• What level of human review is still required for AI-generated production code?
• How do teams prove where code came from and what testing was completed before release?
• Which existing SDLC controls started to break once engineers could produce changes much faster?

Banks have spent years building onboarding controls around documents, selfies, biometrics and customer behaviour. Generative AI is making each of those signals easier to fake just as Australia opens new Digital ID infrastructure to the private sector. 

This keynote looks at how financial institutions are changing identity checks and where Digital ID, biometrics and fraud controls now fit together. 

We'll discuss: 

• Which identity signals can still be trusted when documents, voices and video can be generated?
• Where should Digital ID replace existing KYC checks and where should it only add another signal? 
• How much extra verification can banks introduce before legitimate customers start abandoning the process?

With the AI vendor landscape evolving faster than most institutions' sourcing and procurement processes can track, this Think Tank examines how financial services leaders are making build-versus-buy-versus-partner decisions in a market where today's cutting-edge vendor capability may be commoditised within a year.

• How institutions are managing dependency risk when core AI capability is built on a third-party platform whose direction they can't dictate.
• Where institutions are finding genuine competitive advantage in proprietary AI development versus where off-the-shelf or partnered solutions are the smarter allocation of resources.
• How procurement and contracting approaches are evolving to avoid multi-year lock-in with AI vendors in a market changing this quickly.

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.

As financial institutions move from AI that recommends to AI that autonomously executes; approving loans, rebalancing portfolios, settling claims, this keynote examines what genuinely changes when decisions once made by humans start being made by agents, and sets the tone for a day exploring what that shift demands of governance, culture, and trust.

• Tracing the real trajectory of AI's role in financial institutions over the past few years, and why 2027 marks a genuine inflection point rather than incremental progress.
• The governance, audit trail, and human-oversight infrastructure that needs to exist before an institution should let an AI agent act without a human in the loop.
• Where does agentic AI genuinely create value versus introduce unacceptable risk, and how do institutions build the muscle to tell the difference?

Scam prevention is moving from detection to real-time intervention, with banks increasingly expected to decide whether to warn, delay or block a payment while the customer is still trying to make it. The challenge is getting that intervention right without creating unnecessary friction for legitimate customers.

This keynote explores how financial institutions are redefining payment intervention in the era of increasingly sophisticated scams, including how they set thresholds for action, respond when customers insist a payment is genuine, and navigate responsibility when warnings are ignored.

• What signals and behaviours should trigger a warning, delay or payment block?
• How should banks respond when customers have been coached by scammers and remain convinced the payment is legitimate?
• What happens when a bank identifies the risk, warns the customer, but ultimately allows the payment to proceed?

An AI agent connected to internal systems may have access to customer records, APIs, documents, code and operational tools. Giving an agent too much access can create a much larger problem than an inaccurate chatbot answer.

This panel looks at how financial institutions are thinking about identities, permissions and access controls for AI agents that need to take real actions.We'll discuss:

• Should every agent have its own identity, credentials and clearly defined permissions?
• Which systems should agents never access without a fresh human approval?
• How do you detect when an agent is using legitimate access in a way nobody intended?

As AI models proliferated across the organisation faster than the model risk team could individually review them, one risk leader rebuilt their governance framework around tiered, risk-proportionate oversight, maintaining rigour for high-stakes models without creating a bottleneck for lower-risk use cases.

• How the team identified that applying the same validation rigour to every model, regardless of stakes, was slowing adoption without improving safety.
• The practical criteria used to categorise models by potential impact, and how those tiers determined the depth of review and monitoring required.
• How automated monitoring and documentation tools let a stable risk team keep pace with a rapidly growing model inventory.

A hands-on, interactive session working through a real AI in financial services scenario as a room. Details announced soon.

Facing the challenge of deploying a sophisticated machine learning credit model while maintaining genuine explainability across very different audiences, one lending leader built an explainability layer that satisfied regulatory scrutiny, customer transparency requirements, and legal defensibility simultaneously.

• How the team designed distinct explanation outputs for regulators, customers, and internal dispute resolution, rather than a single generic explanation.
• The trade-offs made between model complexity and explainability, and where the team drew the line.
• How the team stress-tested their explanation framework against real dispute scenarios before deploying the model into production.

After an initial AI-driven claims automation project succeeded in speed but drew criticism for depersonalised handling of complex cases, one insurance leader rebuilt the system to combine automation's efficiency with better outcome measurement and appropriate human escalation.

• How customer feedback and complaint data revealed that speed gains were masking a decline in outcome quality for complex claims.
• How the team redesigned the routing logic to identify which claims genuinely needed human judgement, regardless of dollar amount.
• The shift from processing-time metrics to outcome and satisfaction measures that better reflected genuine customer experience.

AI can help members and customers understand their finances, model options and receive increasingly personalised guidance. The difficult point is deciding when useful assistance crosses into regulated financial advice.This panel looks at where banks, super funds and wealth organisations are drawing that line and how they are designing human escalation into digital advice journeys.We'll discuss:

• What can an AI assistant personalise without crossing into regulated advice?
• Which questions or customer circumstances should automatically trigger a licensed adviser?
• How do organisations scale digital guidance while preserving enough context for a human to take over safely?

Small-group, discussion-based sessions where you'll work through real AI in financial services 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.

AI coding tools are increasing the amount and speed of software produced inside financial institutions. Testing, security review and change controls still have to keep pace with code that may now be partly generated by an agent rather than written line by line by an engineer.This session looks at how one financial institution is using AI across engineering while adapting its review, testing and release controls for a much faster development cycle.We'll discuss:

• What level of human review is still required for AI-generated production code?
• How do teams prove where code came from and what testing was completed before release?
• Which existing SDLC controls started to break once engineers could produce changes much faster?

Banks have spent years building onboarding controls around documents, selfies, biometrics and customer behaviour. Generative AI is making each of those signals easier to fake just as Australia opens new Digital ID infrastructure to the private sector. 

This keynote looks at how financial institutions are changing identity checks and where Digital ID, biometrics and fraud controls now fit together. 

We'll discuss: 

• Which identity signals can still be trusted when documents, voices and video can be generated?
• Where should Digital ID replace existing KYC checks and where should it only add another signal? 
• How much extra verification can banks introduce before legitimate customers start abandoning the process?

With the AI vendor landscape evolving faster than most institutions' sourcing and procurement processes can track, this Think Tank examines how financial services leaders are making build-versus-buy-versus-partner decisions in a market where today's cutting-edge vendor capability may be commoditised within a year.

• How institutions are managing dependency risk when core AI capability is built on a third-party platform whose direction they can't dictate.
• Where institutions are finding genuine competitive advantage in proprietary AI development versus where off-the-shelf or partnered solutions are the smarter allocation of resources.
• How procurement and contracting approaches are evolving to avoid multi-year lock-in with AI vendors in a market changing this quickly.

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