30:55 Lena: Alright Miles, let's get practical here. For our listeners who are convinced that they need to do something with AI but aren't sure where to start, what's the actual playbook? Because I feel like there's a lot of abstract advice out there, but not enough concrete steps.
31:10 Miles: You're absolutely right, and I think this is where a lot of organizations get stuck. They know AI is important, but they don't have a systematic way to move from awareness to actual implementation. So let me walk through what successful organizations are actually doing, step by step.
31:26 Lena: Perfect. Where do they start?
31:28 Miles: The first step isn't technical at all—it's about understanding your current business processes and identifying specific pain points or opportunities. The organizations that succeed with AI start with problems, not solutions. They ask, "What are we struggling with?" or "What would we do if we had unlimited processing power or instant access to information?"
31:48 Lena: Can you give me an example of what that looks like in practice?
20:43 Miles: Sure. Instead of saying "We need an AI strategy," a successful approach might be "Our customer service team spends 40% of their time looking up information in multiple systems, and customers wait an average of three minutes for answers to routine questions." That's a specific problem that AI might be able to help with.
32:10 Lena: So it's about getting granular about the actual work that people do?
0:43 Miles: Exactly. And this often reveals opportunities that aren't obvious at first glance. You might think you need AI to automate entire processes, but often the biggest impact comes from automating specific steps within processes, or providing better information to human decision-makers.
32:30 Lena: What comes after identifying these specific opportunities?
32:34 Miles: The next step is what I call "data readiness assessment." Most AI applications require clean, organized data, and most organizations discover that their data isn't as ready as they thought. You need to understand what data you have, what quality it's in, and what additional data you might need to collect.
32:51 Lena: Is this where a lot of projects get stuck?
32:54 Miles: Unfortunately, yes. Organizations often underestimate the time and effort required to get their data into shape for AI applications. But here's the thing—even if you never implement AI, the process of organizing and cleaning your data usually provides immediate business value.
33:10 Lena: That's a good way to think about it. What about choosing specific AI technologies or vendors?
33:16 Miles: This is where I see a lot of organizations make mistakes. They start with the technology and then try to find applications for it. The better approach is to start with your specific use case and then evaluate which AI approaches might work. Sometimes the answer isn't a cutting-edge AI system—it might be a simpler automation tool or even just better reporting.
33:37 Lena: How do you evaluate whether an AI solution is right for a particular problem?
33:42 Miles: There are a few key criteria. First, is the problem well-defined with clear success metrics? Second, do you have sufficient data to train or configure an AI system? Third, are the stakes appropriate—meaning, can you tolerate some errors while the system learns, or do you need perfect accuracy from day one?
34:00 Lena: What about the implementation process itself? How do successful organizations actually roll out AI solutions?
34:07 Miles: The most successful implementations I've seen start small and scale gradually. They pick a limited pilot project, get it working well, learn from the experience, and then expand. This allows them to work out technical and organizational kinks before committing to larger investments.
34:24 Lena: Are there common pitfalls in the pilot phase?
34:27 Miles: One big one is choosing pilots that are too ambitious or complex for a first project. Another is not involving end users in the design and testing process. AI systems often require changes to how people work, and if you don't get buy-in from the people who will actually use the system, even technically successful projects can fail in practice.
34:47 Lena: What about measuring success during these pilot projects?
34:51 Miles: This goes back to what we discussed earlier about measurement. You need both quantitative metrics and qualitative feedback. Track the specific business metrics you're trying to improve, but also pay attention to user satisfaction, unexpected challenges, and unintended consequences.
35:07 Lena: How do you scale successful pilots to broader organizational use?
35:11 Miles: Scaling requires both technical and organizational planning. On the technical side, you need to ensure that systems can handle increased load and integrate with other business systems. On the organizational side, you need training programs, change management processes, and often updates to job descriptions and performance metrics.
35:28 Lena: Are there organizational structures or roles that support successful AI implementation?
35:34 Miles: The most successful organizations often create cross-functional AI teams that include both technical and business expertise. They might have data scientists and AI engineers, but also business analysts, process experts, and change management specialists. The key is having people who can bridge between technical capabilities and business needs.
35:55 Lena: What about ongoing management and optimization of AI systems?
35:59 Miles: This is crucial and often overlooked. AI systems need ongoing monitoring and tuning. Performance can drift over time as business conditions change or as the underlying data patterns shift. You need processes for monitoring system performance, updating training data, and making adjustments based on user feedback and changing business needs.
36:18 Lena: For smaller organizations that don't have the resources for dedicated AI teams, what's the best approach?
36:26 Miles: Smaller organizations often do better starting with AI-as-a-service solutions rather than trying to build custom systems. Focus on identifying clear use cases and then finding existing AI tools or platforms that can address those needs. The key is still starting with specific business problems rather than trying to implement AI for its own sake.