Stop Building
The Setup
I’ve been having fun building interactive HTMLs (credit to my wife who blew my mind when she showed me something she was playing with). Not as a side project, but as part of how I actually work. Proposals, process maps, frameworks, onboarding guides. Things that used to live in Word docs and slide decks where someone has to read page after page and hope the mental model assembles itself. (Not to mention the dread of spending HOURS assembling it, only to be met with straight confusion afterwards.) The interactive versions let people click, explore, choose their path. And they’ve been landing well. So I started pulling on a thread: what if AI gives us a chance to rethink how we communicate to sync teams’ brains altogether? Not the tools. The actual cognitive moments people have with information at work. That felt like a big enough question to spend a Saturday with. I expected to come out of it with a roadmap for what to build next. Instead I came out with a lesson about when to stop building.
My Approach + AI Role
I used AI as a research partner and we went wide. Scrollytelling, gamified learning, interactive documentation. The engagement numbers are real… 200-300% higher in some studies. But then the counter-signal: a 2024 European workplace study found that higher gamification intensity actually negatively affected perceived autonomy. More game elements made people feel less in control. Then Information Foraging Theory… people forage for information the way animals forage for food. Maximum value for minimum effort. Every click between someone and their answer is a cost. Which means a beautiful interactive HTML might actually be slower than a well-structured Word doc for someone who already knows the material. That research took me somewhere more interesting than I expected. I stepped back from format questions entirely and asked: what are the actual cognitive moments people have with information at work? Seven emerged. I’m still pressure-testing these, and they may end up being six or nine once I’ve lived with them longer. But here’s where they stand:
- Orienting … building a mental model of something new
- Retrieving … finding a specific thing you already know exists
- Sensemaking … holding conflicting signals and forming an interpretation
- Deciding … evaluating tradeoffs and committing
- Producing … making a thing
- Aligning … getting multiple people on the same page
- Verifying … confirming something meets a standard And the realization that made this worth mapping: most organizations serve all seven moments with the same artifact. One doc is expected to orient a new person, serve as ongoing reference, support decisions, AND pass compliance review. That’s why documents are simultaneously too long and not detailed enough. Different moments need different forms.
What Actually Happened
I mapped what I’d already built against the seven moments. Orienting… strong. Retrieving… emerging. Producing… solid but scattered. And then Sensemaking lit up red. The biggest gap. The moment where my personal value is highest… synthesizing conflicting signals, reading between lines, connecting patterns across contexts. Zero tooling for anyone else. The builder-mode kicked in immediately. AI-sensemaking, how might we go about building that? Classic me. I was doing the MBA marshmallow thing. Teams build the tallest spaghetti tower to support a marshmallow. MBA students spend all their time perfecting the structure and stick the marshmallow on at the end… when the weight collapses everything. My architecture was getting more impressive on paper, while the real needs went unaddressed. So I stopped. Wait… am I forcing an AI solution where one isn’t needed? Sensemaking is fundamentally a learning act. When a PM holds three conflicting signals… happy client, burning hours, exhausted team, and works through that tension to form an interpretation and recommendation. That cognitive work IS how they develop judgment. Every time they sit with ambiguity and push through, they build pattern recognition they’ll carry forward. An AI that says “here’s what these signals mean together” without questioning skips the rep for that person building judgment. It automates the exact thing that makes a senior PM worth their rate.
The Real Insight
After I caught myself, TWICE, something connected that I wasn’t expecting. I’ve learned from my wife, an early childhood education expert, that the ability to observe before leaping to interpretation is itself the skill. And it’s the same thing that happens when professionals encounter ambiguous data. The instinct is to immediately interpret. My Builder/AI instinct mirrors it… build a tool that interprets faster. But sitting with observation before leaping to interpretation IS the skill. Don’t automate it. Protect it. I think that’s where AI’s real job becomes clear. Not to think less, but to think with less friction. AI holds (data, memory, context). It surfaces (the right information at the right time). It drafts (initial versions to react to). It does NOT interpret ambiguity, own decisions, or substitute for the social work of shared understanding. The person still journeys through all seven moments. They just move through them with better material.
Try This If
If you’re building AI tools and feeling the pull to automate everything you’ve identified as a problem, pause at the sensemaking bit. Ask whether the ambiguity is actually the training ground where judgment develops. If it is, it isn’t asking for a tool. It’s asking for better raw material and more protected space to think. And try asking your team which of the seven moments they feel most underserved in. Not “what tools do you want” but “when you’re orienting, retrieving, sensemaking, deciding, producing, aligning, or verifying… what makes it harder than it should be?” The answer usually isn’t “I need an AI tool.” It’s “I need the right information at the right time.”
Systems Lens
The most important moments in this whole exploration were the stops. Where I caught myself building a beautiful system and asked whether the biggest gap actually NEEDED a tool. No AI produced that question. It came from sitting with the seven moments and feeling something was off. The pattern recognition that said “this is too clean, we’re doing the marshmallow thing.” My kids’ spring hockey season just started, so my “spend an evening following curiosity” budget is now competing with travel times, rink schedules, and equipment smells that could clear a room. And I think that’s the punch line: The discipline isn’t just about knowing when to stop building. It’s about knowing when to stop researching and go live/sit with what you’ve got. Not everything that’s hard needs solving. Sometimes sitting with it is the work. Funny - I’ve reminded myself of this in a “framework” I’ve seemed to have forgotten back in October here.