Fireside Chat: Operationalizing Agentic AI — Moving from Hype to Business Value
At ROI Training, our mission has always been simple: help organizations bridge the gap between cutting-edge technology and real-world business impact. In this opening session of The Applied Tech Series, our CEO Dave Carey sat down with Google Cloud Difference Maker Trainer of the Year Patrick Haggerty to cut through the boardroom hype surrounding agentic AI and look at what it actually takes to put these tools to work. Together, they unpacked how true AI agents differ from everyday chatbots by breaking down complex, messy workflows into targeted, task-specific applications. They also explored the three tiers of enterprise adoption—from out-of-the-box consumer tools and low-code visual builders to fully custom-coded platforms built with frameworks like the Agent Development Kit (ADK). A recurring theme throughout the conversation was clear: successful AI adoption isn't about rushing shiny tools into production; it comes down to solving actual business problems, strictly curating internal data, and applying traditional engineering rigor to testing.
Beyond the architecture, Dave and Patrick dove into the human side of tech transformation. They addressed widespread fears around job replacement, recasting AI agents as capable mid-level teammates that handle heavy data retrieval and routine grunt work so knowledge workers can focus on high-value judgment, creative strategy, and decision-making. Patrick also broke down practical concepts like "vibe coding"—where human developers step up as senior technical leads while AI handles the boilerplate execution. To wrap up, they tackled the realities of organizational change management. They warned against outright bans on AI tools, which often drive employees toward unsecure personal apps and risk intellectual property leaks. Instead, they shared how proactive teams can deploy secure enterprise environments, build small proofs of concept to prove ROI, and clear backlogs faster than ever.
.png?width=1600&height=900&name=Linkedin%20Live%20Fireside%20Chat%20Graphics%20(3).png)
Key Discussion Highlights
- Defining Agentic AI: An AI agent isn't just a search window; it's a specialized generative application built to tackle specific job roles by breaking down complex problems into manageable sub-tasks.
- The 3-Tier Adoption Spectrum: Organizations implement AI across three levels: standard out-of-the-box tools, low-code visual drag-and-drop builders (like Gems, Custom GPTs, or Claude Projects), and custom-coded platforms built using frameworks like ADK or LangGraph.
- Real-World Impact (UK Bank Case Study): Patrick shared how a major UK bank built a multi-agent system where one agent triaged incoming support emails and a second drafted proposed responses. Human support reps reviewed the drafts split-screen, dramatically cutting down response backlogs while keeping a human in the loop.
- Data Curation & Testing Matter: Hallucinations and awkward outputs usually happen when agents pull from uncurated raw data, like informal internal support notes. AI tools need strict data curation and standard software testing before launch to enforce boundaries.
- Rethinking "Vibe Coding": Effective vibe coding doesn't mean handing full control to AI. The human engineer stays in the driver's seat as the senior architect setting design standards, while the AI acts as a mid-level coder tackling routine tasks.
- Security & Change Management: Simply banning AI tools backfires by pushing employees to use personal devices and free apps, risking IP leaks. Leaders should provide secure enterprise environments and support small, high-value proofs of concept to demonstrate real ROI.