The Artificial Intelligence Show – Episode #171: AI Answers – AI in Regulated Industries, AI Agents, AI Training, Getting It Wrong, and Critical Skills for Early-Career Pros

Hosts: Paul Roetzer (Founder and CEO, Marketing AI Institute & SmartRx), Kathy McPhillips (Chief Marketing Officer, SmartRx)
Date: October 2, 2025

Overview

This special "AI Answers" edition of The Artificial Intelligence Show is a rapid-fire Q&A addressing pressing questions from live classes on AI adoption and scaling. Paul and Kathy draw on their experience teaching thousands of AI practitioners, delivering practical, real-world advice on topics including AI in regulated industries, the difference between generative AI and AI agents, best practices for AI training, what to do when AI gets it wrong, and the most important skills for newcomers in an AI-shaped workforce. True to the show's mission, the episode’s tone is approachable, candid, and grounded in direct organizational experience.

Key Discussion Points & Insights

1. Introducing AI in Regulated Industries (Financial Services, Healthcare, Government)

Timestamp: 06:47

• Main Insight: When using AI in highly regulated contexts, collaboration with IT, legal, and procurement is critical.
• Strategy:
  • Identify risk-free or low-risk use cases (e.g., public podcast content) to build confidence and demonstrate value.
  • Use tools like Jobs GPT to discover low-risk applications tailored to specific roles.
  • "Find ways to infuse AI ... where the risks aren’t there. And then ... solve for the bigger picture." — Paul (07:29)
• Tip: Use AI assistants/chatbots to help outline safe applications and even assist in creating justifications for internal stakeholders.

2. Reimagining Business Models and Customer Engagement with AI

Timestamp: 09:08

• Approach: Use AI reasoning models (like GPT-5 or Gemini 2.5) for brainstorming and strategic planning, not just efficiency.
  • These models can think step-by-step, not just respond instantly—use them like you'd consult a business expert.
• Example: AI-powered chatbots within learning platforms can provide real-time, 24/7 support, freeing human team members for deeper, personal engagement.
• Paul’s Tagline: "What’s more intelligent, what’s more human?" (12:49)
• Anecdote: Even as AI handles more support, make the “off-ramp” to a human easy and pay attention to personalization concerns.

3. Scaling AI: When Organizations Build In-House AI

Timestamp: 15:04

• Five Essential Steps Still Apply: AI education, AI council, responsible AI principles, AI policies, impact assessments, and roadmap.
  • These are ongoing, iterative processes—“impact assessments are an ongoing thing.” (16:16)
  • New AI capabilities (e.g., agentic systems that can write code autonomously) require policies and impact assessments to adapt.
• Change Management: Upskilling and leadership buy-in are crucial. It’s rare for an organization to have these fully solved.

4. Moving from AI as a Side Project to Strategic Priority

Timestamp: 19:28

• Barrier: Lack of education and organizational awareness are the biggest impediments, more than tech itself.
• What Works:
  • Identify who in leadership needs to be convinced.
  • Pilot high-impact use cases; share results—“make it as simple as possible for them to say yes.” (21:59)
  • Build bottom-up enthusiasm by sharing internal success stories.

5. Generative AI vs. AI Agents – What’s the Difference?

Timestamp: 23:02

• Generative AI: Creates text, images, code.
• AI Agents: Take actions, execute multi-step tasks autonomously or semi-autonomously.
  • “AI agents is the ability for AI system to take actions on your behalf.” — Paul (23:09)
  • Examples: Deep research capabilities in ChatGPT or Gemini.
• Policy Implication: Even if not using agents today, organizations should start including them in generative AI policies, as they will become standard features.

6. Shared AI Accounts Among Loosely Connected Teams/Contractors

Timestamp: 26:19

• Practice: It’s technically possible, but each user should ideally have their own account for privacy, personalization, and future knowledge management.
• Warning: Personalization and memory are by user, not centralized—currently, shared organizational “memory” is not unified.

7. AI Guidance for Departments Without an AI Council

Timestamp: 29:59

• Suggestion: Use AI assistants like ChatGPT or Jobs GPT as a “consultant.” Give full context—job role, goals, challenges—to receive tailored advice and use cases.
  • “Getting value out of AI ... comes down to asking good questions.” — Paul (31:55)
• Anecdote: Even middle schoolers (future workforce) are amazed by how AI can tailor advice to niche interests.

8. Best Practices for AI Training and Adoption

Timestamp: 33:15

• Personalize Use Cases: Run workshops tailored to employees’ actual work—create specific prompts or custom GPTs that save them time immediately.
• Executive Participation: If leaders aren’t bought in, customize prompts/GPTs for them to show personal value.
• Addressing Resistance: “Once they see it for themselves, how it can help them ... the lightbulb goes off for everybody.” — Paul (34:51)

9. Driving Engagement When Users Are Too Busy

Timestamp: 35:03

• Tactic: Directly address pain points—identify what’s taking time and automate with custom prompts/GPTs.
• Spark Curiosity: Demonstrate personal and professional applications to encourage organic exploration.

10. Handling AI Mistakes and Hallucinations

Timestamp: 36:14

• Caution: Always keep humans in the loop, especially for high-risk or customer-facing outputs.
  • “The human owns the output ... to own the output, the accuracy and the quality” — Paul (37:32)
• If AI gets company facts wrong: Correct the AI within that session. For team knowledge, upload accurate data into the custom GPT’s knowledge base for broader use.
• Ownership: U.S. copyright law does not grant protection to AI-generated logos and images—serious legal and brand trust considerations.

11. What Skills Matter for Early-Career Professionals?

Timestamp: 40:55

• Top Skills: Curiosity, imagination, critical thinking, communication.
• Educational Advice: Liberal arts skills (e.g., psychology, communication, debate) are key for creative AI collaboration and broader adaptability.
• Entrepreneurship: Dramatically easier with AI—“If I had ChatGPT in 2005, I could have figured out in 48 hours what it probably took me four years to learn.” — Paul (44:10)
• Mindset: Grit and tenacity are as important as tech skills.

12. Ownership and Copyright Risks with AI-Generated Media

Timestamp: 47:20

• Emerging Risk: Major AI models (e.g., OpenAI’s Sora 2) will allow creation of content with copyrighted characters and assets, challenging IP law and brand safety.
• Brand Guidance: Stick to strong moral and ethical guidelines, not just legal minimums.
  • “When legal precedent lags behind, ... your company [needs] a moral compass.” — Paul (48:00)

13. Where Does AI Fall Short? (Expectations vs. Reality)

Timestamp: 50:01

• Main Limitation: Human understanding and change management, not the AI’s current abilities.
• Example: Most organizations don’t realize how advanced AI’s reasoning and deep research projects are, even a year after public release.
• Competitive Advantage: Companies succeeding in change management will gain a strong edge.

14. Ensuring Ethical AI Use and Prompting

Timestamp: 52:06

• Solution: Formalize, document, and actively teach responsible AI principles; make it part of onboarding and daily culture.
• Responsible AI Manifesto: Free for organizations to adapt under Creative Commons. Integrate into GPT knowledge bases for collective referencing.

15. Which Scaling AI Step Is Most Challenging?

Timestamp: 55:05

• Impact Assessments: Rarely done well, but crucial for future-proofing roles, upskilling, and strategic planning.
  • “You have to have a vision for where the AI capabilities are going ... and then from there now you can actually assess jobs and hiring plans.” — Paul (55:05)
• Practical Tip: Regularly run current job descriptions through Jobs GPT to anticipate AI-driven changes.

Memorable Quotes

• “What’s more intelligent, what’s more human?” — Paul (12:49)
• “If you know the questions to ask of a chatbot, you can get a tremendous amount of value with $20 a month.” — Paul (31:55)
• “The human owns the output ... to own the output, the accuracy and the quality.” — Paul (37:32)
• “There’s going to be no excuse to not be able to start a business if that’s what you want to do.” — Paul (46:19)
• “When legal precedent lags behind, ... your company [needs] a moral compass.” — Paul (48:00)

Timestamps for Key Segments

• [06:47] – AI in Regulated Industries
• [09:08] – AI for Innovation, Stretching Beyond Efficiency
• [15:04] – Advanced Organizations & In-House AI: Principles and Change Management
• [19:28] – Making AI a Strategic Priority
• [23:02] – Generative AI vs. AI Agents
• [26:19] – Shared AI Accounts: Pros, Cons, and Privacy
• [29:59] – Getting AI Guidance Without an AI Council
• [33:15] – Training Best Practices for New AI Users
• [35:03] – Driving Busy Team Member Engagement
• [36:14] – Handling AI Mistakes and Hallucinations
• [40:55] – Most Important Skills for Early-Career Talent
• [47:20] – Copyright, Brand Risks with AI-Generated Content
• [50:01] – Where AI Falls Short (It’s About Us, Not AI)
• [52:06] – Building Ethical AI Practices and Prompting
• [55:05] – The Toughest Step: AI Impact Assessments

Overall Tone:
Supportive, pragmatic, occasionally humorous, with an emphasis on real-world strategy rather than theory. Both hosts share technical advice as well as personal anecdotes, making the complex world of AI feel accessible and actionable to organizations at all levels.

For more: Check out the podcast’s show notes for links to referenced tools (e.g., Jobs GPT, Responsible AI Manifesto), upcoming events, and workshops.