← All essays
6 min read

Killing Product Zombies: A Leader's Guide for the AI Era

SM
Swanand Marulkar

Your product portfolio is hiding a secret that is killing innovation: the product zombie. It is the legacy feature on life support, the pet project with ghost-town usage, the platform that is not dead, but is not truly alive.

A few years ago, I laid out a framework for killing product zombies. So why am I writing about this again?

The current LLM boom has fundamentally rewritten the playbook for product development. On one hand, it allows us to create new products and features at unprecedented speed, but this very speed creates a new, more insidious breed of zombie: features without strategic rigour that quickly bloat our products. On the other hand, many ambitious AI projects become PoC zombies. They get stuck in a perpetual R&D cycle, always one model away from the perfect solution, consuming vast resources without ever delivering value to a customer.

The true cost of inaction

Before we can act, we have to understand the real damage a zombie inflicts. The cost is not just a line item on a budget. It is a drain on the entire organisation, and in the age of AI, this drain is faster and deeper than ever.

  • The financial cost: the most obvious and consistent bleed. Beyond direct maintenance and accumulating technical debt, AI-powered zombies introduce new, unpredictable operational expenses. A feature that relies on an external API can suddenly become a massive financial drain through variable token costs, turning a popular feature into an unprofitable one overnight.
  • The opportunity cost: the strategic price you pay. Every dollar and every engineering hour spent keeping a zombie alive is taken from your future. This is dangerously amplified when your scarcest and most expensive talent, your AI and data science teams, are tied up on a perpetual PoC zombie instead of your most critical business opportunities.
  • The human cost: the one leaders most often ignore. Nothing burns out a great team faster than forcing them to maintain a product they know has no future. It quietly kills morale, motivation, and the innovative spark you hired them for.
  • The brand and reputation cost: your brand is a promise of quality and focus. A zombie product dilutes that promise, sends a confusing message to the market, creates a disjointed customer experience, and forces your sales team to explain away the past instead of selling the future.

The LLM boom is forcing us to look at the original framework again, and at how the mindset, finding and evaluation of zombies change in the AI era. A successful triage is not a single action but a disciplined, strategic process. It requires a clear-eyed assessment before any decisions are made.

Step 1: The mindset in an age of easy creation

The fundamental mindset for this era is strategic toughness. In the past, the high cost of development acted as a natural filter for weak product ideas. Today, with powerful AI tools, the cost of creating a new product or feature, the initial output, has plummeted. This age of easy creation is intriguing, but it creates a dangerous trap: it makes it easy to confuse activity with progress.

The only way to avoid the trap is to double down on the fundamentals of good product management. The core principles do not change. They just become more critical.

This requires a tougher, more disciplined focus on outcome over output. The output is no longer just a design mock-up. It can be a fully functional prototype created in a single afternoon, which makes the output feel substantial and creates a stronger emotional attachment. The modern discipline is having the courage to look at an impressive AI-generated product and say: this is a technically brilliant output, but it fails to deliver the one customer outcome that matters.

A powerful, ongoing example is Meta's pivot to the Metaverse. Announced with a company-wide rebrand and backed by tens of billions of dollars, the output was immense. The customer outcome, however, has been a well-documented failure. One analysis highlights that after burning over $60 billion on the initiative, Meta's flagship platform, Horizon Worlds, attracted fewer than 200,000 monthly active users (Magnus919, 2025). It is a stark reminder that even unlimited resources cannot guarantee value, tying up capital and talent that could be deployed on more immediate opportunities.

Step 2: Finding the product zombies

The original principle of this step is timeless. Finding zombies requires a deep investigation that goes beyond raw data, relying on the gut feeling of your product team, honest conversations with sales and support, and, most importantly, listening to your customers. In the AI era, with so much noise and hype, these human-centric investigation skills matter more than ever.

Consider the Apple Vision Pro. A technological marvel, yet a true investigation means listening to your customers, in this case early adopters and developers. The qualitative feedback since its 2024 launch has been consistent: a feeling of isolation, a lack of a killer app, and physical discomfort during extended use. Investigating these human signals, not just initial sales figures, is critical to understanding the potential zombie aspects of the product strategy and finding a path to true market fit.

Step 3: Triage

Be ruthless and realistic. What would a genuine revival effort cost, and what is the evidence-based, projected ROI? Is a massive reinvestment the most valuable thing you could do with those resources, or is it throwing good money after bad?

After its failed IPO, WeWork's new leadership had to perform a massive triage. They could not afford to cure every unprofitable location. They had to assess the ROI of each individual office lease, deciding which to keep and which to abandon, to stop the financial bleeding and create a viable business.

This is the critical step for today's PoC zombies. Triage forces a ruthless question: are we truly one model away from a breakthrough, or are we throwing good money after bad in an endless R&D cycle with no clear path to production?

Steps 4 and 5: The final decision and execution

After triage, the process converges with the original framework. The final steps, decision and execution, are principles of leadership that technology does not change. You must still make the tough call: Cure, Contain, or Decommission. And if the path is to decommission, as Warner Bros. Discovery did with CNN+, the execution must be swift and clear, providing a respectful end for a product's life.

The five-step framework is not a magic wand. In a world where the vast majority of new features fail, zombies are an inevitable part of the product lifecycle. The framework's true power lies in turning the messy, political process of cleanup into a disciplined, data-informed practice. Today, that discipline is augmented by two force multipliers.

  • Pragmatic AI and data: you do not need a massive data science team. You need simple, automated dashboards that spotlight reality faster, and tools to quantify the real cost of support. This is not about predicting a perfect future. It is about using data to have honest conversations in the present.
  • Empathetic leadership: data can show you what to do, but it cannot convince your team why. Facilitation, leading a team to a shared conclusion, managing stakeholder emotions, and communicating a clear, respectful plan separate a difficult decision from a destructive one.

Ultimately, this is a pragmatic act of reclaiming your most valuable resources: focus, capital and energy to fuel your next great innovation.

What is the biggest product zombie lurking in your portfolio right now, and what is stopping you from decommissioning it? If you want to talk it through, book a conversation below.