artificial intelligencemarketing
Why longevity and adaptability after deploying agentic AI will define enterprise success in 2026
Spokesperson: Adam Beavis, Country Manager Australia and New Zealand, Databricks.
Q1: Agentic AI has moved quickly from experimentation to deployment. What will separate organisations that succeed from those that fall behind after rollout?
A: What separates organisations that win with agentic AI after rollout from those that stall is less about the tech — and more about the operating model and discipline. The key differentiators tend to be:
1. Data readiness: High performers invest heavily in clean, permissioned, continuously improving data and instrument agents with feedback loops. Without this, agent performance degrades quickly after initial rollout.
2. Strong guardrails and governance by design: Winning organisations bake in controls, auditability, escalation paths, and human-in-the-loop thresholds from day one. Those that fall behind treat governance as an afterthought—leading to trust issues, halted deployments, or regulatory friction.
3. Clear business ownership, not just tech ownership: Successful firms tie agents to specific business outcomes (cost, speed, risk reduction, revenue uplift) with accountable business unit owners.
4. Cultural and behavioural change: The biggest gap is human, not technical. Leaders who succeed redesign roles around human–agent collaboration, and retrain employees to integrate AI into their daily work and oversee autonomous systems.
Q2: Many organisations feel they have done AI once it is deployed. What often goes wrong for organisations beyond that point?
A: The biggest misconception is treating deployment as a box-ticking exercise. Models that are trained on historical data can drift as inputs change, and without consistent and continuous evaluation, problems often surface too late.
The solution is shifting from one-off checks to continuous evaluation in production. Just like humans need performance reviews, so do AI systems. Enterprises need systems that continuously measure performance against real tasks, retrain or adjust agents and balance quality against cost. Many early deployments struggle, because they were not designed with long-term operation in mind.
Q3: Why is the transition from single agents to multi-agent orchestration important for enterprises?
A: Enterprise work rarely happens in a single step. A realistic workflow often includes retrieving from multiple data sources and validating data against business rules, compliance checks, and a final decision with explainability requirements. Expecting a single agent to handle all these tasks reliably and efficiently is unrealistic.
In 2026 we will see broader adoption of multi-agent orchestration, where specialised agents handle distinct tasks and a supervising agent coordinates sequences, mirroring how human teams operate. While this gives the benefit of better performance, it also allows for improved governance. Each agent can be monitored and evaluated on its specific responsibility, modifications can be isolated, and the overall system remains transparent and auditable, and easier to troubleshoot.
Q4: Many enterprises struggle to get AI agents and applications into production. What is causing the bottleneck and how is Databricks addressing it?
A: The bottleneck is not building a demo, it is making agents reliable in the real enterprise. In productions, agents must consistently reason over complex, proprietary data, operate with guardrails and integrate with operational systems.
General knowledge of AI is becoming a commodity, but it’s still elusive to get AI that truly understands the proprietary data inside an enterprise. Many AI agents fail in enterprise environments because they prioritise ease of use over accuracy, leading to inconsistent results or behaviour organisations cannot trust.
Databricks is addressing this by building through a number of growth areas:
- The rise of AI-powered coding is changing how software is built. As developers create apps via natural language processing, those apps automatically need databases and agent backends. Databricks is seeing this first-hand, with over 80% of databases launched on Databricks now being created by AI agents, rather than humans. We are enabling developers to rapidly build applications that run on Lakebase and are powered by agents, all within a unified and governed platform..
- Enterprises need a modern transactional layer for AI-native apps. Traditional transactional databases have changed little for decades, so we launched Lakebase, which simplifies operational data workflows and is optimised for AI agents operating at machine speed.
- Agent Bricks then helps organisations build and deploy agents that can securely work within their own data, where most of the business value sits. It helps organisations build domain-specific agents that reason over their data, track quality with task-specific benchmarks and balance performance with cost over time. The aim is to make agent quality measurable and improvable in production, not assumed at deployment.
Together, this trifecta removes the common production blockers by combining an operational database layer, an agent-building platform that works with enterprise data and an application layer that helps ship faster, with reliability and governance built in.
Q5: In Australia and New Zealand, how are organisations specifically adopting AI applications and how does that differ from earlier phases?
A: Across Australia and New Zealand, we’re seeing a pivot from general-purpose experimentation to domain-specific AI applications that are grounded in trusted enterprise data, with stronger attention to governance and sovereignty. As organisations shift from pilots into production to deliver real business outcomes, organisations are now embedding AI into real workflows, from customer support and supply chain to finance and operations.
For example, Suncorp needed to scale AI across the organisations to improve claims accuracy, reduce operational risk and support more automated digital customer experiences. Manual processes were creating additional loads for staff, and employees often lacked instant access to complex policy and claims information when decision making was required. By building, deploying and scaling domain-specific AI directly into claims workflows with Databricks, the company has achieved 99% accuracy and saved more than 15,000 hours of manual workload.
Additionally, Atlassian uses Databricks AI/BI Genie to power on demand insights in plain English through Atlassian Rovo, its AI assistant. This allows teams across the business to ask complex questions of their data and receive trusted, contextual answers directly within their existing workflows.
These examples reflect a broader trend that we’re seeing across the region in 2026, where value increasingly comes from AI applications designed around the business, grounded in governed data, and operationalised end to end.
Q6: How are AI applications, including AI agents, playing out across key verticals? What advice would you give to enterprise leaders in these sectors?
A: Whether you lead in finance, pharma, media, CPG, or tech, the questions are converging: how do we use AI to improve business productivity? How do we balance industry regulation with AI innovation? How do we control costs without slowing adoption? The leaders who solve these challenges today will build faster, more resilient operations and gain a competitive edge.
In the public sector, AI is connecting data across agencies to reduce administrative burden and support decisions from benefits assessment to emergency response. Success depends on strong governance, clear lineage and transparency so outputs can be trusted and audited.
In marketing, AI applications are moving beyond content generation to orchestrating campaigns, analysing performance data and adapting system strategies in near real time. Data Intelligence for Marketing allows organisations to centralise customer and campaign data, apply AI to drive more accurate decisions, and automate tasks that scale human resources using AI agents.
In cybersecurity, multi-agent systems are proving effective to validate threats and accelerate response times while keeping humans in the loop. Databricks’ Data Intelligence for Cybersecurity powers scalable SecOps at scale with Agent Bricks by automating triage, enrichment, response and investigation, reducing alert fatigue and costs while boosting analyst productivity.
My advice for leaders is simple:
● Invest in your data and AI foundations with high-quality data and governance
● Have clear business ownership and outcomes you want AI to accomplish
● Scale what is already working.
