AI is no longer a side conversation inside the enterprise. The question has shifted from “Should we invest in AI?” to “How do we operationalise it in a way that is useful, governed, and measurable?”
That was the thread running through Tealium’s recent Architect Arc events in Sydney and Melbourne, where we brought together data architects, engineers, platform owners and AI leaders to explore what it really takes to move from AI experimentation to production-grade execution.
Having been in the room for these sessions, and presenting as part of the series, what stood out most was how quickly the conversation has matured. There is still plenty of excitement around AI, but the focus has clearly moved beyond hype. People are now asking sharper, more practical questions: How do we connect the right data to the right models? How do we make AI useful in a live environment? And how do we do it in a way that is governed, scalable and grounded in business value?
Across both cities, the discussion centred on a common challenge: getting AI into production. The blocker is often not the model itself, but the infrastructure behind it — the ability to connect context, consent, decisioning and action in real time. That is where Tealium’s role came into focus throughout the series, as a real-time context layer for AI that connects customer touchpoints, data clouds, AI platforms and downstream engagement systems with governed, actionable data in motion.
Three practical AI patterns stood out
One of the strengths of Architect Arc was that it grounded the AI discussion in practical patterns that teams can apply now.
The first was AI connector-style decisioning, using prompt-based and low-code integrations with services such as Amazon Bedrock, OpenAI, Anthropic and Vertex AI. This pattern helps teams move quickly by connecting AI services directly into operational workflows without needing to rebuild everything from scratch.
The second was Invoke Your Own Model (IYOM), which gives organisations a way to use their own models for real-time scoring and recommendations while keeping those models in their own cloud environment. For teams that want more control over architecture, governance and performance, this offers a practical path from experimentation to activation.
The third was real-time context for AI agents. This is where the conversation became especially compelling. AI agents become far more useful when they can retrieve profile data, interpret signals, update state, respect consent and trigger actions across systems as part of the same interaction. In other words, intelligence alone is not enough. AI becomes valuable when it is connected to live customer context and operational systems.
From presentations to hands-on working sessions
Architect Arc was intentionally built as more than a speaker series. The day moved from presentations into live demos and then into hands-on working sessions, giving attendees a clearer view of what it takes to operationalise AI in practice.
Having seen the demos, and witnessed the art of the possible, it was illuminating to hear what topics were top of mind for attendees in those workshop sessions. They ranged from the functional – “we’ve got a chatbot, but it only knows about our products and services, not the customer it is interacting with” – to the evaluative “how do we know that we can rely on an AI decision?”, “how do we govern AI agents?” – to the pragmatic “what can we allow the agent to decide to do vs what should we prescribe as a process that is deterministically followed?” and “when is it safe to release an agent into production?”.
And there were lots of use case ideas, like “do we think this customer has a pet?”, “make a personalised newsletter subscription offer”, “is this customer about to churn?”, “create the perfect bundle of products for this customer”. Attendees left enriched by the discussion and shared experiences, with practical ideas for putting their AI use cases into action and we look forward to helping drive successful outcomes.
Another big part of that experience was seeing Tealium capabilities brought to life through live demonstrations, including Tealium Mobile SDK. Rather than talking about AI and real-time customer context in the abstract, these demos showed what it looks like when customer signals are captured, stitched together and activated in ways that support more responsive, context-aware experiences. It made the conversation feel much more immediate because people could see the mechanics, not just the messaging.
The live-build session added another layer of depth. I led a practical walkthrough on how to build an agent, showing how real-time data, orchestration and AI workflows can come together in a usable architecture.
Seen as one of the highlights of the day, it gave attendees the chance to move beyond theory and see how these ideas actually come together. Instead of leaving with abstract inspiration, teams could connect the dots between architecture, workflows and real-world use cases they could take back into their own environments.
Bringing AI and customer context to life
Another standout element was the Architect Arc Café app, which gave attendees a live, customer-style experience of Tealium capabilities in action. The app supported coffee and food ordering, agenda access, the AI Accelerator offer and AI-powered food pairing recommendations that combined on-device AI, cloud AI connectors and rules-based fallback logic.
It also pulled live visitor profiles via Moments API, with PRISM handling event capture, identity stitching and the full data layer.
What I particularly liked about this part of the experience was that it made the broader story feel tangible. Instead of talking about real-time data and AI in theory, attendees could interact with an experience that showed how these pieces work together in practice. It turned the day’s central message into something people could actually see and use.
More importantly, it reinforced a simple point: AI delivers more value when it is connected to identity, consent, orchestration and real-time action rather than operating in isolation.
The takeaway: AI needs a real-time context layer
If there was one idea that connected the Sydney and Melbourne events, it was this: organisations do not just need AI tools. They need a real-time context supply chain that helps models and agents understand who they are interacting with, what matters right now, what they are allowed to do and how each outcome should inform the next decision.
That is the opportunity Tealium is helping customers unlock, whether they are starting with AI connectors, operationalising existing models through IYOM or building agentic experiences powered by real-time customer context.
What I left both events thinking was that the market is moving quickly, but also more thoughtfully. The appetite is there, but so is the need for practical architecture, governance and clear paths to value. For the teams that joined us in Sydney and Melbourne, the next step is not more theory. It is follow-up: refining the use cases identified during the day, moving the best ideas into live evaluation and turning AI ambition into production outcomes.
If your team is focused on turning AI ambition into execution, explore Tealium for AI to see how real-time customer data can power more relevant, governed AI experiences.
