Navigating the AI-Driven PDLC: How Product Teams Are Adapting

AI isn’t just changing how teams write code; it’s reshaping the entire product development lifecycle (PDLC). As coding velocity accelerates, the bottlenecks shift upstream: to decisions, alignment, dependencies, and portfolio governance.

These are challenges most product teams are actively navigating right now. The build side is accelerating, but the processes, operating models, and structures around it weren’t designed for this pace. The pressure is showing up in real time: in portfolio calls that are harder to make, alignment that’s harder to hold, and dependencies that are harder to track.

Join us for a fireside chat between Hakan Yilmaz, Director of Product Portfolio Operations at Advisor360°, and Becky Flint, Founder & CEO of Dragonboat, as they dig into what’s actually shifting across the product lifecycle and what it means for how teams plan, align, and operate – sharing candid takes on what’s working and what’s still a real challenge.

You’ll hear:

• Why faster code doesn’t mean faster decisions
• Navigating AI as both a product feature and as a development tool
• The new challenges in portfolio decisions, dependency management, and alignment
• Enabling your product teams in this new AI era

Product Ops HQ is brought to you by Dragonboat, the product operating system that helps teams and AI agents decide, plan, and act across the full PDLC—aligning strategy, execution, and outcomes at AI speed and scale.

Key Takeaways

The Three-Pronged Approach to AI Adoption

Hakan outlines how Advisor360 approaches AI adoption across three distinct areas of their business:

• Prong 1: Within the Product Itself
  The team began by introducing chatbot-style tools to help their wealth tech users with specific, individual tasks.

  “We started by introducing individual chatbot type agents for a particular workflow or particular task to gain efficiency for the end user.”
  However, they quickly realized that chatbots are only as good as the underlying data they access. To expand beyond basic helpers, Hakan notes: “you need to have an underlying, fundamentally sound data structure that agents can sit on top of and talk to.”

• Prong 2: Within the PDLC Execution
  At the execution level, tools like Claude and Augment allow engineers to code and test at unprecedented speeds. Rather than using this newly created capacity to simply write more code, Hakan leverages it to bring developers into early-stage strategy.

  “Using things like Augment or Claude… engineers can code faster and do more testing is easier… what we can also do is take the same time they’ve saved and pull them earlier into the PDLC process… so they’re doing more discovery amongst themselves.”

• Prong 3: Operationalizing Across the Organization
  The final stage is moving from individual use cases to scaled, company-wide workflows.

  “It’s actually enabling the entire organization to adopt AI as a company, not just at an individual level but at an organizational level—so implementing collective workflows that we can all use and all benefit from, going from AI assistance to AI augmentation to being AI native.”

Shifting the PDLC Focus: Frontloading Discovery and Prototyping

Because AI has compressed downstream coding timelines, the organizational bottleneck has naturally shifted to early-stage decision-making, requiring a complete rethink of how product teams operate.

• The Rise of Small Prototyping Pods:
  To experiment safely without disrupting critical production environments, Hakan advocates for starting small:

  “We’re adjusting the whole PDLC and frontloading a lot of the work and emphasizing the importance of judgment and discovery… doing that in smaller teams initially so engineer, designer, product manager, one or two others seeing how those pods work and going from there.”

  “We’re spinning up some small teams, just teams of five people… making sure we’ve got a really nice prototyping mechanism, we’ve got engineers involved up front and seeing what they can come up with and that way we’re experimenting quickly.”

• Engineering Involved in Ideation:
  Giving engineers a seat at the table during the discovery phase ensures the team makes better early-stage architectural and product decisions:

  “I’ve long been a proponent of having engineering involved in ideation very early in the process… the advent of AI gives us the opportunity to do that very well.”

Taming the Chaos: Process and Data Infrastructure as a Prerequisite

When asked what advice he has for companies whose processes are chaotic and who fear AI will only accelerate that chaos, Hakan emphasizes establishing a tool-backed foundation and process discipline before writing any prompts.

• Establishing the Portfolio Tooling Foundation:
  Hakan explicitly highlights Dragonboat as his primary tool of choice to build a stable organizational framework:

  “Have a great portfolio management tool at your disposal like Dragonboat, I do that. But beyond that I think what you want to do is take a step back from AI for a second and just look at the fundamentals of product management, your PDLC. Make sure that’s well defined.”

• Defining the Process Gates First:
  A portfolio tool is useless if the product organization does not have basic operational hygiene:

  “Just make sure those [gates] are adhered to, make sure you’ve got executive buying executing on those, and then you can layer in how AI can augment and help gain efficiencies. You want to make sure you’re defining your ideation phase… how do they get added to the roadmap, what are the gates there, how do we agree, what are the metrics…”

• Feeding the Decision-Making Engine:
  Ultimately, Hakan notes that internal decision-making data must be curated with the same rigor as customer-facing LLM data to avoid generating what he calls “eloquent nonsense”:

  “In the same way that you need data to feed an LLM model for your product, you need data to feed the product decision-making engine.”

  “Don’t just lump agents up there. If you have agents without the data infrastructure, you just get eloquent nonsense.”