Skip to content
← All essays

AI

Context Rules Everything

“Malcolm often says in 7 words what could be said in 1.” — a piece of feedback I received earlier in my product career. I was more frustrated with this when I received it, but now I laugh because…it’s wrong.

And to clarify, the person who gave me this feedback feels this way, and his feelings were valid. However, my takeaway wasn’t so much that I needed to be concise as it was that he didn’t have the right information to understand the value of the additional 6 words. In other words, he was lacking context.

At another point in my career, I was working with a particularly challenging executive, and they wanted an update on some of my work. I shared the update (about 7 bullets…something about that number) with my leadership in advance, and pretty immediately she said, “Oh. Let’s get this shorter. He’s not going to read all of this.” But I also knew this executive preferred context to be shared up front. So I didn’t respond right away. After a few minutes had passed (the time needed to actually read it), she came back, “Wait. He needs all this information. Let’s just send this.” Could I have trimmed the 7 bullets to 1? For sure. But the context of the last 6 would’ve been missing, which would’ve only resulted in more back-and-forth, wasting time and money for everyone involved.

In my time at People, I experienced “3x3s” for the first time: employees email 3 points of progress and 3 opportunities to a broad set of stakeholders weekly. Everyone has their own list, and the more senior you are, the more you are summarizing across a wider portfolio of work, which is perhaps obvious. The challenge each week was to be as succinct as possible while also providing enough relevant context so the information presented could be properly interpreted. (“I would have written a shorter letter, but I did not have the time.” was often quoted—and misquoted.) The challenge was heightened by the fact that each email is sent to tens (sometimes hundreds) of people, and each individual receives tens of them… each week.

I would have written a shorter letter, but I did not have the time.

– Blaise Pascal

Most of the emails I received in this fashion had little value because most people considered them a frustrating, tick-the-box task each week. And, unfortunately, that is a vicious cycle, a self-fulfilling prophecy. The primary function of these emails was to effectively communicate strategy and vision out to organization members who are not as close to your work, but most used them to provide status updates that would only be meaningful to someone closer to the work. Again, there was a dearth of context.

And now, it seems many have woken up to the value of context (and its management), because it’s easier to perceive when working with AI. It shows up in LLM outputs to the point that compact and clear commands are critical for day-to-day workflows. Context management also has a clear impact on token usage, meaning there is a stronger through-line to P&L, which has made people more willing to face the concept of context head-on.

As AI tooling becomes more ubiquitous, we’re circling back to what I’ve known all along: context rules everything. Without it, we work less efficiently, make less impactful decisions, and are more prone to mistakes. As I wrote last week, information architecture is a critical skill for today’s PMs (and anyone scaling LLM usage at their organization). This is due, in large part, to the fact that this understanding will lead to better context management, which, in turn, will yield a positive, attributable impact on P&L. Because it’s never been just about providing context in a vacuum. It’s about providing the right amount of context, at the right time.