AI and the Context Problem - An Opportunity for Product Management
- Grant Elliott
- Mar 7
- 3 min read
Updated: Mar 28
There's been a lot of discussion about AI’s impact on the role of product managers—some even predict that AI might replace PMs altogether. As I’ve written before, it’s essential to distinguish between two very different roles often lumped together:
Product Delivery Managers (PDMs) – Focus on execution, delivery timelines, and process management.
Traditional Product Managers (PMs) – Own the vision, strategy, and problem-solving that drive a product’s success.
AI will significantly reduce the role of PDMs, automating much of the execution-heavy work they manage today. But for traditional PMs, AI represents not a threat—but a massive opportunity.
There are many reasons for this, ranging from ethical considerations (who ensures AI makes responsible product decisions?) to functional limitations (who guides AI’s output toward strategic goals?). In this article, I want to focus on one of the most immediate and pressing AI challenges: the Context Problem.
The Context Gap in AI
AI has evolved rapidly, but it still struggles with one critical element: context.
Unlike humans, who naturally build and apply broad contextual understanding, AI models like GPTs perform well in narrow contexts (e.g., a single conversation or document) but struggle when required to synthesize information across wider bodies of work.
This limitation is both a challenge and an opportunity—especially for product managers.
The Limitations of AI: Why Context Breaks Down
AI’s power is often overstated, leading to déjà vu moments where we think it's about to replace human expertise, only to hit fundamental walls.
GPTs and LLMs can:
Write solid code within a single function but struggle to design an entire software architecture.
Generate a document but fail to integrate it meaningfully into a long-term product strategy.
Understand a single-threaded conversation but lose coherence when pulling from multiple sources over time.
Why this happens:
Token limitations mean AI operates within a constrained memory window.
No true persistence. Without structured data storage, AI forgets previous sessions.
AI lacks hierarchical reasoning, making it weak at structuring long-term context across multiple interrelated elements.
Silicon Valley’s Mistake: Overestimating AI, Undervaluing Product Managers
In "Did Silicon Valley Kill Product Management?", I argued that the PM role has been diluted by a feature-first mentality rather than a strategic approach to product development. The AI hype cycle is repeating this mistake assuming AI will be able to self-manage complexity rather than requiring structured, human-driven context design. The reality? AI isn't replacing strategic decision-making—it needs PMs more than ever to frame, structure, and provide context.
The Product Manager’s Opportunity: Designing Context-Aware AI
Product Managers must bridge AI’s context gap by designing products that embed and retrieve context effectively. This includes:
Structured Knowledge Retention: Using embeddings, vector databases, and linked data to allow AI to retrieve historical context dynamically. this requires understanding the purpose and the intent of the application, something often lost to the development team.
Context-Aware AI Design: Building AI-driven features that consider long-term product vision and not just short-term queries. This goes back to the "why" of the product and can be as much 'art' as it is 'science'.
Human-in-the-Loop Feedback: AI alone won’t get context right. Product managers must curate, refine, and validate how AI integrates information. As I have written before when discussing the dangers of synthetic data, product management must play a critical role in maintaining the integrity of synthetic data.
AI’s Future Needs Context, and Context Needs Product Managers
The AI revolution won’t be a simple déjà vu of the internet or cloud revolutions. It requires new thinking to solve for context. AI won’t kill product management; it will make it more essential. The best products in the AI era will be those that effectively embed context. And that’s a problem only great product managers can solve.
Comentarios