AI Answers: Navigating Strategy, Adoption, and the Future of Work

In this edition of AI Answers, Paul Roetzer (Founder and CEO of SmarterX and the Marketing AI Institute) and Kathy McFillips (Chief Marketing Officer at SmarterX) tackle a series of pressing questions from business leaders and practitioners. This episode focuses on the practical realities of scaling AI, managing organizational change, and separating AI hype from actionable strategy.

Amazon and the Maturity of AI Rollouts

Question: Is Amazon slowing its AI rollout a sign of maturity?

Paul suggests that recent reports of Amazon slowing down parts of its AI deployment—often due to issues with "rogue" agents—likely reflect the "growing pains" inherent in rapid innovation. Rather than a calculated sign of maturity, it is a reality check for the industry. Large-scale models and autonomous agents are advancing faster than our ability to perfectly secure them. This creates a messy environment where "experimental speed" must be balanced with "responsible safety." The takeaway for businesses is to avoid rushing into agentic workflows without robust human-in-the-loop oversight.

Are Large Enterprises at a Disadvantage?

Question: Are large enterprises structurally disadvantaged in the AI era, or do they have assets that allow them to win?

Paul categorizes businesses into three groups:

While large enterprises face significant hurdles—such as inertia in service-based pricing models and outdated software stacks—they possess critical advantages: massive customer bases, deep talent pools, and financial strength. If leadership can embrace change and avoid the "innovation trap," they can outpace smaller competitors. However, many current C-suites are struggling to pivot with the required urgency, often leading to turnover and valuation shifts.

Who Owns the Data and Adoption Bottlenecks?

Question: Who owns the problems of AI adoption and data readiness inside an enterprise?

The reason these issues remain unsolved is that many organizations treated AI as a pure IT project. By assigning the responsibility solely to the CIO, companies failed to educate business unit leaders (CMOs, sales heads, etc.) on how to integrate AI into their specific workflows.

Furthermore, "data readiness" is often a red herring used to delay adoption. Paul notes that roughly 90% of initial AI use cases—particularly in marketing—do not require access to proprietary internal data. Businesses can and should begin their AI journey with projects that don't touch sensitive data while cleaning their infrastructure in parallel.

The Growing "AI Divide"

Question: Is there a growing AI divide between power users and others?

Yes, and it is a major organizational problem. In any team, you will find three groups: those who actively resist AI, those who use it for surface-level tasks (e.g., meeting summaries), and "power users" who are racing ahead.

Power users become exponentially more productive, effectively performing the work of multiple peers. The hard reality for leaders is that employees who refuse to use these tools may eventually become obsolete. Paul recommends a human-centered approach: provide clear expectations, offer personalized training, and give employees a runway to adapt. If they still refuse to use the technology after receiving the necessary tools and training, they may no longer be a fit for an AI-forward organization.

The Future of Jobs and Automation

Question: What is an AI take you have that most people disagree with?

Paul remains concerned about the long-term impact on employment. While he believes AI is a net positive for the economy, he anticipates a challenging period of both unemployment and, more importantly, underemployment.

Regarding the balance between automation and augmentation, Paul suggests a shift based on seniority:

The "No-Brainer" Use Case

Question: What is an AI use case that feels like a no-brainer but most companies haven't tried?

The most valuable, underutilized application is the Strategic Thought Partner. If you aren't using the advanced reasoning models (like the latest versions of Claude, Gemini, or ChatGPT) to help brainstorm strategies, solve complex problems, or critique your logic, you are leaving massive value on the table.

Paul suggests a high-value workflow:

  1. Run the same prompt through three or four different AI models.
  2. Choose the best output as your "primary" thread.
  3. Use the other models to critique and refine that primary output.

This process might take a few hours, but it replaces weeks of manual labor and significantly improves the quality of the final result.

Strategic AI Governance

Question: Where is governance necessary, and where does it get in the way?

Governance is non-negotiable when a use case involves:

If an AI agent is empowered to email a customer or alter a database, you need guardrails. However, governance often becomes bloated and kills innovation. Paul recommends creating a "Center of Excellence" or a smaller subcommittee that can test and experiment in a sandbox without being slowed down by the bureaucracy of the entire organization.

Final Advice for Leaders

Paul emphasizes that the most common mistake in AI strategy is the failure to start with AI literacy. You cannot effectively manage, budget for, or delegate an AI strategy if you don't personally understand the technology's capabilities.

His final message to those struggling to get started: Start small. Don't wait for a mandate from the top. Find one painful task—like a Sunday night reporting chore—and build a custom tool to automate it. Once an employee experiences the "life-changing" benefit of offloading a mundane task, they will naturally become an advocate for further experimentation.