Reflections from the Melbourne Enterprise AI and Automation Summit 2026

1. Governance and security are the defining constraint

One of the biggest barriers to deploying AI programs within organisations is not model capability – it is governance and security. Data sovereignty, IP ownership, explainability, vendor risk and regulatory exposure are front of mind. In many cases, the technology works. The challenge lies in aligning risk, legal, IT and business functions before anything goes live.

Many organisations are building their governance frameworks while simultaneously deploying AI tools, and that is arguably just what adoption looks like in practice. Organisations are catching up to AI in real time, iterating as they go rather than waiting for a perfect policy that may never arrive.

Risk tolerance ultimately shapes the framework. Highly regulated or risk-averse organisations tend to design layered, conservative controls. Others prioritise agility. What became clear is that governance must be fit for purpose, there is no single AI policy that can be copied and pasted across sectors. There is also a concern that overly heavy governance may slow innovation. The challenge is designing guardrails that enable experimentation rather than suppress it.

Underneath all of this sits a quest for confidence and certainty. Organisations want to move forward, but they want to do so knowing the foundations are solid.

2. Trust takes time, so start early

AI cannot be bolted into an organisation and expected to gain acceptance. It helps to embed trust from the outset through transparency, education and clear communication about how AI is being used, what data is involved and what controls are in place.

Without that groundwork, adoption tends to stall internally. Without it, customers ask harder questions. And once trust is compromised, rebuilding it costs significantly more than investing in it upfront.

3. The cost of adopting, and the cost of not adopting

A recurring tension was the dual-sided cost equation. Implementation requires real investment: infrastructure, security, governance, training and change management. But the cost of not adopting AI is becoming increasingly visible. Competitors are embedding AI into workflows, reducing delivery times and reshaping margins.

This creates genuine strategic pressure. Delay too long and market position may erode. Move too aggressively and the organisation may take on operational or regulatory risk it is not prepared to manage.

Some organisations are now facing customer scrutiny around AI usage in contracts. Clients are asking whether AI is being used in service delivery and, if so, whether efficiency gains should translate into lower pricing. The AFR recently reported that KPMG demanded a discount from its own auditor on the basis of AI-driven cost savings, a signal that AI is not just a productivity lever but an emerging commercial and contractual variable.

4. Getting from pilot to enterprise

Many AI initiatives do not struggle at the proof-of-concept stage, they struggle at scale. Embedding AI into core systems, workflows and decision-making processes requires clean data architecture, clear ownership, defined ROI metrics and genuine cross-departmental alignment. Without those foundations, AI projects risk remaining isolated experiments.

The statistic cited at the summit, that around half of AI projects fail, reinforces the point. The failure is rarely about the algorithm. It is almost always about execution discipline.

5. AI is reshaping operating models, not just tools

Perhaps the most important theme was this: AI is not a software upgrade. It is an operating model shift.

Deploying AI changes how work is performed, how decisions are made, how risk is assessed and how value is priced. It affects structure, capability and culture. Treating it as a discrete technology project underestimates its impact considerably. The organisations that appear most prepared are those thinking holistically: process redesign, talent capability and governance architecture in parallel, not in sequence.

Enterprise AI is as much about structure, risk and trust as it is about the technology itself. The pace of change is accelerating, but the frameworks governing that change are still very much being written.

It will be interesting to see how this evolves.