Practical lessons from the afternoon sessions: governance engineering, AI security, executive perspective, and semantic layers. 

ADG Summit is Australia’s biggest event for Data and AI Governance, run by DGX. On 20 August in Melbourne, we brought together 200+ delegates, more than 25 speakers, and our headline, gold and industry sponsors to share what works in the real world. 

This is Part 2 of the recap, covering the afternoon program.
If you missed it, Recap Part 1 covered the morning sessions, from Actian on real metadata architectures to keynotes and panels on agent governance, building demand for governance, trusted data, practical AI governance, and privacy engineering. 

After a packed morning, the room stayed full, and every afternoon session cut straight to implementation. Here’s what stood out. 

Keynote: The Case for Governance Engineering 

With Andrew Hinds (Data Technologist, CDMP Master) 

Rising expectations, falling budgets and diseconomies of scale mean manual governance cannot scale. Andrew’s answer is governance engineering, the discipline he pioneered at Bupa: engineering skills and technology applied to governance processes. Data isn’t the new oil, he argued, it’s the new uranium. 

Key insights 

• Governance engineering: engineering technologies and skills applied to governance processes 

• Delivered so far: classification at ingestion, masking driven by that metadata, policy as code in a data product marketplace, lineage across 100% of the platform, golden record as a service, agents monitoring data quality 

• Governance engineers are made, not found: pair a technical governance person with a governance-minded engineer 

• You don’t need a restructure to start, but you do need influence over policies and standards 

• Buy or build is the wrong question, the answer is both 

• Mindset is the anchor, and the reframe is enablement over policing 

Standout quote 

“We need to not be seen as a handbrake on the business, but the highly performant brakes of an F1 car.” 

Andrew Hinds 

Panel: AI Security: Governing Agents and MCPs 

With Reshma Devi (Transurban), Blare Sutton (McGrathNicol), Vinod Bijlani (Hewlett Packard Enterprise). Moderator: Bernadeth Lucanas (AI Governance Consultant) 

What changes when agents can act and MCP gives them access to the enterprise. Deliberately not a technical session: the focus was the governance lens, what has actually gone wrong, and which existing controls still hold. Reshma’s question to the room, how many agents does your organisation have, produced almost no hands. 

Key insights 

• Perimeter security still helps and does nothing about attacks arriving inside the data 

• AI adds a cognitive layer: intent inspection, non-human identity, guardian agents 

• Data-aware guardrails checked at the point data is returned from an MCP call 

• Before autonomy: foundations, guardrails people know about, automated and tested controls, monitoring on a cadence matched to impact 

• The dangerous failures are agents behaving as designed but outside intent 

• Most first agents are pointed at unstructured data with no labelling or catalog at all 

• One organisation identified roughly 1,400 staff-created agents on a single platform 

Standout quote 

“We don’t want to fall in love with the solution. The focus will always be what is the business case, and we should weigh the opportunity against the risk that comes with it.” 

Bernadeth Lucanas 

Fireside: Data Governance: An Executive Perspective 

With Siva Sivakumaran (UniSuper) and Tim Moon (DGX) 

Ten months into holding both strategy and data, Siva explained how governance gets explained to a board: not in the language of the discipline, but as three risks. Security, quality and privacy. The key risk indicators follow from those, and there are four or five of them, not hundreds. 

Key insights 

• Strategy and data belong together: strategy should be driven by data, not thought bubbles 

• Three risk buckets for the board: security, quality, privacy 

• Keep KRIs to four or five, and keep the deck to three slides 

• Narrow the critical data elements, make executives the owners, have them name the stewards 

• Tier stewardship, because not all stewards are created equal 

• Retention and disposal is a no-brainer: when it works nobody notices, when it stops everybody asks 

• Focus on the outcome, not the process or the journey 

Standout quote 

“I struggle to read more than three slides. If you can’t tell me something in three slides, you’re hardly likely to tell me something in ten.” 

Siva Sivakumaran, UniSuper 

Panel: Human Centred Governance: Starting With People, Not Frameworks 

With Darshil Mehta (AustralianSuper), Karina Blackman (GrainCorp), Trisha Lee (Vanguard). Moderator: Rina Gami (Konnexus) 

In 2012 the first question on a governance project was which framework to implement. Now it’s how this helps people use data safely and correctly. The panel’s test: ask the business six months after launch how your framework helped them do their job better, and see whether you get an answer. 

Key insights 

• Change the narrative about yourselves: be the face of governance and a person before a framework 

• Make it personally relevant at every level: regulation for execs, time saved for middle management, daily work for operations 

• Co-design standards with the people who have to follow them 

• Tailor change management to each business, not generic slides 

• Start your answer with yes, then explain how to do it safely 

• Crowdsource shared problems, because more people with the same problem means funding 

• Human at heart: automate the repetitive, keep humans on interpretation and accountability 

Standout quote 

“Compliance shouldn’t be a stick. We should always try to start with the carrot first.” 

Trisha Lee, Vanguard 

Panel: Semantic Layers Demystified 

With Ole Olesen-Bagneux (Actian) and Joshua Yu (GraphWay AI). Moderator: Nikolai Zorin (DAMA Melbourne) 

Ask three data people to define a semantic layer and you get four answers. The panel separated glossary, metadata, ontology and semantic layer, then explained why decades-old technology suddenly matters again: agents need a precision that humans used to work around. 

Key insights 

• Four different things: a glossary defines terms, metadata describes assets, an ontology is a structure of meaning, a semantic layer is an architectural capability 

• The ontology is the bridge from intellectual structure to technical implementation 

• Semantically, customer is a role an entity plays, not a record in a table 

• Expect multiple semantic layers coupled to specific agentic architectures, not one enterprise layer 

• Ownership scales with autonomy: thin for chat over documents, cross-functional for autonomous processes 

• Start from the glossary, reference model and catalog you already have, then link and extend 

Standout quote 

“Please don’t start at the whiteboard. The existing knowledge of the semantic universe in the enterprise already exists, so find it before you recreate it.” 

Ole Olesen-Bagneux, Actian 

DGX Data & AI Governance Awards 

At the ADG Summit 2026 we announced two awards, presented by Ole Olesen-Bagneux from Actian, our headline sponsor. 

The Data & AI Governance Leader of the Year went to Phillip Ward, Senior Manager, Privacy Engineering at Canva. Phil was recognised for building privacy engineering into a discipline that makes compliance the default path for engineers instead of a policy they have to interpret. He also delivered a morning keynote on low-debt data governance. 

New this year, the Rising Star in Data Governance went to Brad Hill of Aware Super, recognised for leading-edge work using AI to govern AI. Aware Super also produced last year’s inaugural Leader of the Year. 

Congratulations to both, and thank you to our finalists, whose work is shaping the future of Data & AI Governance in Australia. Nominations come from the community, so keep it in mind for next year. 

That’s it for our ADG Summit 2026 recap. 

A huge thank you to our speakers, sponsors, industry partners, and every delegate who joined us in Melbourne to make it such a success. We look forward to continuing the conversations and to seeing you at ADG Summit 2027.