← all posts site ↗

balancing strategic ai work and tactical mvps

i'm in this weird position at work where i'm supposed to be building a system for intake and prioritization of ai initiatives across the whole commercial banking division. it's a big strategic thing that involves finance, tech, transformation, and risk orgs. but right now everyone's attention is on a one-week qa tool mvp that a teammate is shipping. the qa tool is sexy because it's a demo people can see. my system is invisible because it hasn't shipped yet. the irony is not lost on me — the very friction my system is meant to eliminate is what's driving current priorities.

the core tension is role definition. i'm not a developer but i'm not a manager either. i'm this connective tissue thing between what gets built and whether the org actually gets smarter about ai with each initiative. when my teammate demos her tool everywhere and gets pitched as the ai builder, it's hard not to feel like my work is being deprioritized. but the fix isn't competing for the same kind of visibility. it's naming a different role entirely. i'm the architect of the enablement system. she's the builder shipping fast. both matter but they need different kinds of airtime.

the move that made the most sense was to let her own the sprint and i own the retro. while she builds i extract the lessons. what blocked us, what governance friction came up, what the build speed actually says about our enablement readiness. that artifact becomes the template for every future initiative. i also need to score the five inbound build requests sitting in her pipeline through my prioritization framework today, not later. score them against actual bank imperatives and create something shareable — a simple list she can link people to when they ask can you build x. suddenly my invisible system has a url and a purpose.

the bigger reframe is that demos win short term and infrastructure wins long term. both are real. the trick is making infrastructure visible before it's done. a one-page retro, a scored pipeline, a stakeholder signal summary — these are artifacts that can be shared now, not when the agent ships. you attach your system to the thing that already has leadership attention without claiming credit for what someone else built. you become the person who ensures the org gets smarter with every initiative, not just the high-profile ones. the qa tool isn't a distraction from my work. it's the first proof point for why my system needs to exist.