When Does It Make Sense to Actually Use AI?
Episode Overview
Agentic AI is everywhere, but does every problem actually need an agent? In this special edition of Hot Takes, Zack Austin and Fran Fernandez bring in Nelson Veiga, Resolve's Senior Vice President of Customer Solutions, to react to Gartner's When to Use or Not to Use AI Agents report. Drawing on real-world enterprise implementations, the trio explores why organizations are shifting from AI experimentation to measurable business outcomes, where deterministic automation still has an advantage, and how enterprises should decide when autonomous reasoning is actually necessary.
Key Takeaways
- The AI conversation is shifting from adoption to outcomes. Organizations were once under pressure to invest in AI simply to demonstrate that they had an AI strategy. Now, though, boards and executives increasingly want to know what those investments are actually delivering, putting greater emphasis on quantifiable business value.
- Agentic AI doesn't replace deterministic automation. Autonomous agents can reason through variable situations, but execution still requires a reliable layer capable of taking action. The real opportunity comes from combining AI reasoning with deterministic automation instead of treating the two approaches as competitors.
- Not every workflow needs autonomous reasoning. Predictable processes like account unlocks, service fulfillment, and access provisioning already have known paths from request to resolution. Introducing unnecessary variability can increase complexity and risk when traditional automation can execute the same work consistently.
- You can delegate authority, but not responsibility. An autonomous agent may execute work, but the organization deploying it remains accountable for the outcome. Higher-risk use cases therefore require clear guardrails, transparency, auditability, and appropriate human oversight.
- Human approval will evolve alongside autonomous AI. Traditional approval workflows often give reviewers limited information before execution. As agents become more capable, organizations may need higher-fidelity oversight that shows what an agent plans to do (or what it’s done) so human operators can act accordingly.
FAQ
Q: How should enterprises decide when to use agentic AI versus traditional automation?
A: Start with the use case rather than the technology. When a process follows predictable steps and requires consistent execution, deterministic automation may be faster, safer, and easier to audit. Agentic AI becomes more valuable when the work genuinely requires reasoning, adaptability, or the ability to navigate greater variability.
Timestamp: 17:50–24:10
Q: Who is responsible when an autonomous AI agent makes a mistake?
A: The organization deploying the agent remains accountable. As Nelson puts it, companies can delegate authority but not responsibility. That makes transparency, governance, security, and human oversight especially important when agents are allowed to take actions that could affect sensitive systems, data, or access.
Timestamp: 22:07–24:10
Q: How should organizations prepare for AI becoming more autonomous and accessible?
A: Nelson recommends establishing an AI center of excellence or similar governing team responsible for evaluating tools and deployments. As powerful AI capabilities become increasingly accessible to individual employees, organizations will need stronger oversight to prevent shadow AI from introducing security, compliance, and operational risks.
Timestamp: 39:00–43:46
