Published on the 25/11/2025 | Written by Heather Wright

The year of AI and agentic execution…

AI hype is giving way to hard work, with Deloitte predicting 2026 will be the year businesses shift from flashy pilots to the gritty reality of scaling and enterprise integration.

Getting in early with their year-end predictions, the professional services company has released its Technology, Media and Telecommunications Predictions report, signalling a clear move from flashy demos and proof-of-concepts to the gritty reality of enterprise adoption. With that though, comes a clear challenge for IT leaders: How to turn those AI pilots into scalable, compliant and value-driven solutions.

“Initial uses of AI agents to support people and automate tasks can drive three to five percent in annual productivity improvements at a company level.”

“The gap between promise and reality will narrow, but not disappear,” the report says, noting that progress will come less from headline-grabbing new models or even shiny new enterprise agentic applications (though, let’s be honest, they will still continue to garner attention) and more from fundamentals – the data hygiene, integration into existing workflows, governance, new pricing models and regulatory compliance, required to translate AI beyond pilots and trials.

AI adoption across Australia and New Zealand has surged in the past two years, though clarity on return on investment is another story. Reports from Datacom, SAP, Cisco and Adobe all agree businesses are investing heavily in AI, but offer widely differing accounts of the payoffs businesses are experiencing. Adobe’s 2025 AI and Digital Trends report claims A/NZ has seen the fastest AI adoption growth in APAC in the past year, with adoption doubling from 14 percent to 29 percent. But despite that just 12 percent of local brands report consistent returns on investment (up from zero percent in 2024).

SAP, meanwhile, says Australian companies are already achieving 15 percent ROI, and forecasts that figure to nearly double by 2028. (Their spend figures, however, are eye-watering, with the report saying Australian organisations are yielding an average ROI this year of US$3.2 million on an average US$19.1 million spend.)

Cisco, meanwhile, warns that fewer than one in five New Zealand organisations are fully ready to capture AI’s potential.

Datacom’s State of AI index hints at the reasons for the disparity, noting that while 87 percent of Kiwi businesses use AI in some form, only 12 percent have scaled it across their enterprise.

Whatever the figures, the underlying message is the same: Scaling AI is hard and the gap between ambition and execution is wide.

That’s also clear in Deloitte’s report.

From solo to symphony: Agent orchestration

Among the report’s 13 key topics is one on unlocking exponential value with AI agent orchestration. The company says agent orchestration – or the effective coordination of role-specific agents – will be essential to unlock the potential of agents, with poor orchestration significantly limiting business value.

Claims about agentic AI’s value are, like those for AI in general hard widely differing and hard to verify.

A recent McKinsey report says early implementations suggest there is ‘significant’ value at stake, with McKinsey’s own experience modernising tech estates indicating that agents can accelerate timelines 40-50 percent and reduce costs more than 40 percent while improving quality of outputs.

“Our experience indicates that the initial uses of AI agents to support people and automate tasks can drive three to five percent in annual productivity improvements at the company level,” McKinsey says. “As teams of AI agents become capable of carrying out more complex workflows, growth could increase by as much as 10 percent or more.”

At the IT Symposium/Xpo on the Gold Coast earlier this year, Gartner senior director analyst Anushree Verma suggested more than 40 percent of agentic AI projects will be cancelled by the end of 2027 due to unclear business value, escalating costs or inadequate risk controls.

Verma told attendees most agentic AI projects were early-stage experiments or proof of concepts driven largely by hype and often misapplied, blinding organisations to the real cost and complexity of deploying AI agents at scale and stalling projects from moving into production. She warned that many vendors were engaging in ‘agent washing’, rebranding existing products such as robotic process automation, AI assistants and chatbots, even though they didn’t have substantial agentic capabilities.

“[Organisations] need to cut through the hype to make careful, strategic decisions about where and how they apply this emerging technology.”

A Gartner poll of nearly 3,500 webinar attendees early this year saw 19 percent saying their organisation had already made ‘significant’ investments in agentic with 42 percent having made conservative investments. Just eight percent hadn’t made any investments and the remaining were taking a wait and see approach.

Verma’s advice to IT leaders was to only pursue agentic AI where it delivers clear value or ROI.

Integrating agents into legacy systems can be technically complex, often disrupting workflows and requiring costly modifications. In many cases, rethinking workflows with agentic AI from the ground up was the ideal path to success.

““To get real value from agentic AI, organizations must focus on enterprise productivity, rather than just individual task augmentation.

“They can start by using AI agents when decisions are needed, automation for routine workflows and assistants for simple retrieval. It’s about driving business value through cost, quality, speed and scale.”

Deloitte, meanwhile is calling for enterprises to orchestrate agents better and address the challenges and risks head on in order to drive agent success.

It says companies should consider three potential multiagent approaches: A smart overlay, agentic by design or process redesign.

While adding AI agents on top of existing, well-defined workflows enables quick experimentation and may reduce legacy system disruption, its pitfalls can include challenges around integration, cost control and data security.

Agentic by design, which sees processes restructured to apply new custom-built AI agents for select workflows, can lower barriers to implementation, but can also see innovation, security and compliance potentially tied to provider capabilities.

Process redesign, for high-priority areas where automation is difficult and risk, sees processes ‘rewired’ and agents deployed, but while it can unlock novel and innovative use cases, Deloitte notes it also may require careful planning and phased execution.

Research also shows today’s emerging multiagent systems continue to perform better with humans in the loop.

“We predict that, in the next 12 to 18 months, more businesses will accelerate experimenting and scaling of complex agent orchestrations, keeping humans in the loop,” Deloitte’s report says.

“They will likely adopt frameworks and solutions to integrate human judgement into agentic workflows for higher confidence, quality and accountability.”

But, it says, 2026 will see the most advanced businesses beginning to lay the foundations for the next step in the evolution of AI agent autonomy: Humans on the loop – where agents gain greater authority to make decisions, but check and collaborate with humans to review outcomes.

There’s also a warning for businesses, with Deloitte saying agentic AI will make SaaS even more complex, as vendors build in AI agent-powered offerings to existing products, create agent-building frameworks on top of current services and introduce new data management and orchestration capabilities to make the creation and management of AI agents easier, while new AI-native companies offer solutions which could potentially disrupt incumbents.

More complex markets like ERP and CRM are likely to be disrupted, albeit not immediately.

While organisations are likely to initially take an ‘agentic-by-default’ approach, adopting agents from their existing providers, over time they will likely shift toward a more deliberate tack, with Deloitte predicting companies will pick and choose capabilities from a broad and complex agentic ecosystem, develop their own agents and weave everything into an integrated and autonomous multi-agent system.

The company is also forecasting that the user experience and interface for SaaS AI agents will undergo a transformation, becoming more personalised, proactive, conversational and diagnostic.

Companies using SaaS offerings need to invest in data management and help workers become AI orchestrators managing AI agents like co-workers. “It’s a cultural shift, not just a software upgrade.”

And, Deloitte says, it’s time to embrace the growing complexity of multi-faceted pricing models and more agents, vendors, ecosystems and data relationships.