Contextualizing context
I have never resonated much with the metaphor of PMs as CEOs of the product. While it is a useful call to the agency and leadership aspects of our roles, it tends to imply far more centralized power than I believe should exist in a good product team - and more than actually exists for most PMs. But recently, I have started likening the PM role to a different C-level position that doesn’t exist (yet) - PMs as the Chief Context Officers of the product.
As the idea of context management becomes more central to building with & using AI, it feels to me that the PM role is naturally well-suited to leading the charge. Even in a pre-AI world, a core responsibility of the PM role has always been to elicit, manage, curate and convey product context to a variety of constituencies. So, while the more novel and revolutionary ideas of “universal context” and “company brains” are gaining traction nowadays, it is worth pointing out that us PMs have a ton of reps already acting as the “universal product brains” for our organizations.
I come across a lot of discussion and writing these days that focuses on how best to manage context. From mastering prompt engineering, to debating whether Markdown or HTML is the optimal format for your context files, to optimizing agent harnesses like IDEs in the before-times, to investments at organizational levels in universal context that LLMs & agents can benefit from as much as humans - these ideas tend to focus on how context should be managed. But I see a lot less written about what we actually mean by context i.e. the what, not the how. My intent for today’s essay is to balance that out a bit, by sharing 5 key characteristics of what I believe good and useful context actually is. I hope this helps all of you budding Chief Context Officers, and your teams, to form a stronger mental model of context.
1. Context is infinite
The universe of potentially useful context is practically infinite. Even if you are talking about a specific decision or task at hand, the context that could be useful to action it with confidence can - theoretically at least - exist anywhere and in any form. Useful context can be buried in the Slack thread from years ago when we last talked about this issue, to the CEO’s strong opinion (loosely held?) in last quarter’s strategy review, to your brilliant shower thought over the weekend, or maybe the competitor launch announcement earlier this morning that just shook things up. That doesn’t mean all potential context is useful; 99% of it is often noise. It just means that there are no preordained, hard edges to your universe of context. It also means you need to draw useful context boundaries yourself. And, like the colonial overlords of eras past discovered, the act of drawing boundaries is inherently judgment-driven and opinionated - and comes with tradeoffs.
Implication: context management needs deliberate curation, not just aggregation. You need to have an opinion about what the most relevant and useful context substrate is for your team or org, and to understand the tradeoffs of what you choose to edit out.
2. Context is perpetual
Useful context - similar to useful code, or design - has a past, present and future to it. For example, any meaningful product decision we make today is simultaneously:
A culmination of past insights and intuitions, and a continuation in a lineage of past product decisions
A representation of a specific intuition or opinion we want to act on now
A shadow cast on future results, learnings and future decisions yet to be defined
Context is not a snapshot, it is evergreen in nature. And as we now attempt to systematically manage perpetual context, we will unlock newer opportunities (ex: decision logs/traces that compound over time into Context Graphs), while needing to contend with newer problems (ex: context rot, since context tends to organically atrophy in usefulness over time). Institutional memory needs to be treated as a first-class asset, not a passive artifact that will accrue and manage itself.
Implication: managing perpetual context is substantively different from maintaining a historical record or decision-log in a database. It is more like building a useful institutional memory bank, with necessary abstractions and change management designed into it.
3. Context is nuanced
Subtle shifts in how we manage context can produce meaningfully different outcomes. Many of us think of context as a set of objective truths or fact-base for all of us to work off of. In practice, product context is often a mix of facts and opinions, of reality and our perceptions of it. That’s why even subtle changes in the language we use to capture context, or the density of detail we decide to capture, or the subjectivity we allow into it can produce different interpretations. “Oh, that one detail changes everything!” is a dynamic us humans, especially in small numbers, are better equipped to deal with than AI agents - for whom real-time nudges, nuance and course-correction can be terribly expensive.
Strong PMs radically own the nuance that is inherent to product context; they thrive in it. It is why they are able to seamlessly connect and communicate around it with a diverse set of stakeholders, to dynamically move up/down in altitude and abstraction depending on the need of the hour and (often frustratingly :)) produce multiple artifacts that can make the same underlying fact-base about a product more accessible and useful to different contexts or audiences. This isn’t just work-about-work; it’s often the critical work to orchestrate convergence and momentum through a diverse, complex organization.
Implication: embrace the fact that context is a mix of facts and opinions, of objective truth and our subjective understanding and perception of it. Don’t succumb to the temptation that all useful context can be flattened in a simplified, templated manner - even as we seek out more accessible context.
4. Context is negotiated
As we talk about the implications of context being infinite, and perpetual and nuanced in nature, there is an elephant in the room worth addressing. Context must be prioritized and curated, to be managed. And as all of us product people know, prioritization is fundamentally a negotiation across multiple parties and multiple needs and multiple constraints.
Consider a deceptively simple question: “is initiative X on track?”. This feels like the type of question that can be reasonably answered with a factual answer of project status that can be sourced from a centralized tool. That’s true - but in most organizations, there are a series of implicit questions & concerns that are actually baked into & around a general question like that, ex:
What is the real goal or deadline for initiative X? (note: not always the stated one)
What is the quality bar for the work being done as part of initiative X? Is it hitting that bar?
What are the tradeoffs being made? Are these the right tradeoffs?
What if the initiative looks like it’s ‘on-time’ but because we cut scope that was valuable?
What risks do we foresee? What are we doing about those risks?
Are prospective users vibing with the value prop or idea? Why? Why not?
What is the health of the team working on X?
The true answer to how initiative X is going is a mix of both the fact-base i.e. project status along with a dynamic, amorphous concoction of negotiated answers to questions like the ones above. The true answer is negotiated and emergent, before it can be ratified. It requires active mind-melding of the brain trust involved in the work - not just output from the “company brain”.
Implication: context management needs an active system of negotiating a shared reality. Consider the collaborative and convergence-building forums, rituals and vocabulary around the raw context - not just the output.
5. Context is cultural
You can take the context out of the culture, but you can’t take the culture out of the context. The way an organization creates, captures, collates and conveys context is a deeply cultural act, even when implicit. It reflects hierarchies (ex: whose context counts, and whose gets filtered out before it reaches the room). It reflects collaboration and communication norms (ex: is context shared proactively, or hoarded? is it centralized and standardized, or decentralized and fragmented?). It reflects psychological safety (ex: is the uncomfortable and real context brought to life, or does it fester outside of the new systems?) And it reflects organizational and institutional memory (ex: to what degree to do we learn from history, relative to the present or future?).
Context management presents human questions, not just technical questions. Universal context feels egalitarian in intent - but its truthfulness and effectiveness will ultimately be gated by the quality of the organization’s leadership and the organizational culture. This is because companies are actually more like continuously evolving ecosystems, than singular bodies with a single brain.
Implication: Invest in the cultural conditions that make good and useful context possible, not just the technical ones. The ceiling for the impact of your context management system will be culture, not tooling.
A parting note, for us “Chief Context Officers”
The 5 characteristics above are not a checklist - they are a lens to help you understand the criticality and complexity of good context. In a world where organizations increasingly need systematic context management, product leaders are best positioned to be the executives of that change, and to influence how it happens. And now, you have a stronger mental model for what you actually need to manage.
Wrap
Have a question or musing of your own?
Feel free to drop a comment below, or DM me.
Enjoyed today’s musing?
Consider sharing with a colleague or friend who may benefit too.
Interested in more of Waqas’ Musings?
Find all my past essays at https://waqas-sheikh.com, and subscribe below 👇