Knowledge architecture
Eight groups, no shared system for what they know, turned into a governed operating model
For Knowledge Manager / KM Analyst roles
A media company’s operations and technology division: eight groups, dozens of subgroups, no shared system for its knowledge. Here is how I build one from scratch, discovery to a governed steady state, with an AI agent doing the drafting.
I ran discovery and set the structure across all eight groups, designed the readiness model the program worked from, and designed and prototyped an AI intake agent that drafts governed docs. The full migrate-and-operate cycle was piloted on one group before the engagement ended, and the gap it exposed became All Clear.
- 01
Discover & structure
Made a complex org legible
Resolved an unconventional media-tech structure into a verified map of groups, subgroups, and champions, and cleaned up the shared sites that had drifted into disorder. Roughly two working sessions per subgroup, across all eight groups. The org map and champion matrix gave leadership one place to see where the whole program stood; the information-gathering canvas was where each subgroup’s champions dropped in what was ready to share.
Interactive artifacts
- 02
Assess
Designed the readiness model
A five-stage scale, Dreaming to Exemplary, that made every team’s progress toward a governed knowledge state visible at a glance and gave dozens of stakeholders a shared language.
Interactive artifact
- 03
Migrate
A governed move, not a free-for-all
An AI-assisted inventory of each site (metadata only), a routing decision made by the team’s own champions, then an automated relocation of every asset to its right home, with access verified before anything moved.
Interactive artifact
- 04
Operate
Designed an intake agent for the steady state
A knowledge intake-and-governance system whose centerpiece is an AI agent: it drafts knowledge in the right shape, enforces the fields that matter, routes for approval, files into an access-controlled library, and never fills in what it has not been told. I built and demoed the working prototype.
Interactive artifact
- Dreaming
- Collecting
- Organizing
- Ready
- Exemplary
The problem the intake agent kept surfacing became its own project.
When the engagement ended, I took the core problem, knowledge that is technically saved but not trustworthy or findable, and built All Clear, a working prototype that tackles it directly.
Every artifact above is rebuilt on a fictional company so the method shows without exposing anything real. The five stages, the readiness model, the agent design, and the pilot are how the work actually went.
Open to a senior full-time role. Hiring brief →