Process Automation with AI Agents - How to Think Before You Build | | Zero Shot Ep. 24

The foundations most people are working from are already out of date. Here is how to think about automation now. If you have ever built an automation, watched someone demo one, and thought you should be doing more of this, this episode is for you. James Gauci (CEO of Cadent) and Llew Jury (MD of Advancer) go back to first principles on process automation in 2026. Not a tool tour that is stale in six months, a durable mental model you can carry as the tools keep changing underneath you. The big shift is this. Automation is no longer about structured data and fixed rules. The move from deterministic if this then that systems to probabilistic reasoning systems changes what is possible, what is worth building, and how you should think about applying it, even in regulated and compliance heavy industries where people assume it is too hard. In this episode you will learn: 🔀 Deterministic vs probabilistic systems, the real shift under the hood and why it matters 🧩 Why automation is no longer about clean structured data and rigid rules 📊 Data lakehouses explained, why getting your data in order comes before any agent 🏛️ Automation in regulated industries, how finance, health, and aged care can do this safely 🧠 The judging problem, why the part you assumed needed a human is now automatable 🛡️ The human in the loop vs the meat shield, the honest framing nobody else gives you 🎯 The four Ps, problem, people, process, product, and why tool selection comes last 🔮 What separates the businesses that get this right from the ones still building the old way ⏱️ Timestamps 0:00 , Welcome to Episode 24: Process automation from first principles 2:38 , Sponsor: Prompt Cowboy 3:16 , Tool of the week: running local models for private on-device inference 6:30 , How we got here, from rigid rule-based automation to reasoning systems 9:41 , Adding reasoning, the real break from 20 years of if this then that 12:58 , Why businesses get stuck on the old model, legacy, cost, and knowledge 21:00 , The shift that changes everything, data lakehouses and unstructured data 28:18 , What messy data capability changes about what a company can do 30:07 , Finance agents and real-time insight, from slow reconciliations to live 30:55 , Automation in regulated industries, data sovereignty and onshore control 36:33 , The trap, automating the low-value part because it is easy to picture 40:13 , The four Ps framework, why product and tool selection come last 45:29 , Key takeaways: collapse the right processes, not the obvious ones 🔗 Resources and Links 🔌 Zapier , Workflow automation platform, the original if this then that tool for connecting apps: https://zapier.com 🗂️ Airtable , Flexible relational database and automation layer for small business workflows: https://airtable.com 🎙️ OpenAI Whisper , Free open-source speech-to-text model that runs locally for private inference: https://github.com/openai/whisper 🔗 Model Context Protocol , Open standard for connecting AI agents to your data and tools: https://modelcontextprotocol.io 📐 ISO 42001 , International standard for AI management systems, a governance starting point: https://www.iso.org/standard/81230.html 🎧 Spotify: https://open.spotify.com/show/76JaUrD... 🔗 LinkedIn:   / zero-shot-ai-podcast   🌐 Zero Shot: https://zeroshot.com.au/ 🎙️ About Zero Shot Zero Shot is the anti-hype AI podcast for business leaders, executives, and entrepreneurs who want the technical and commercial truth about artificial intelligence. Hosted by James Gauci (CEO of Cadent) and Llew Jury (MD of Advancer), and produced by Yennia La Rotta. We cut through the noise to bring you the strategies that actually move the needle. New episode every week. Subscribe so you never miss the filter. 🎙️ 💬 What is the messiest high-value process in your business right now? Drop it in the comments. James and Llew read every single one. 👍 Like , 💬 Comment , 🔔 Subscribe Process automation in 2026 is not about clean data and rigid rules anymore. The shift from deterministic systems to probabilistic reasoning means AI can now read messy, unstructured information, make judgements, and produce structured outputs that used to require a human. From data lakehouses and the Model Context Protocol to automation in regulated industries and the four Ps framework, this episode gives business leaders and decision-makers the first-principles thinking they need to automate the right things safely and for real return. If you want a mental model that outlasts the tools, this is where to start. #ProcessAutomation #AgenticAI #BusinessAI #ZeroShotPodcast #AIStrategy #DataLakehouse #Automation #AIAdoption #FutureOfWork #ArtificialIntelligence #AIForBusiness #ModelContextProtocol #WorkflowAutomation #FirstPrinciples #Australia