Roll Cognition out to your team without the mess
Going from one person to a whole team is where shared-knowledge tools usually rot into a landfill nobody trusts. Here is the rollout that scales the brain while keeping it governed, private, and actually trusted, with the signals to aim your effort.
Shared knowledge bases tend to die the same way. Everyone is told to contribute, nobody is responsible for curating, and within a month it is a landfill: stale entries next to current ones, duplicates everywhere, and a quiet team consensus that searching it is not worth the time. Once that trust is gone it almost never comes back. Cognition is designed to prevent that specific failure, but only if you roll it out in a way that keeps the safeguards on. Here is that rollout, in order.
1. Start with a brain, not a blank page
An empty tool on day one invites the worst behavior: people dumping everything in to make it feel useful. Avoid that by starting with a prebuilt starter brain for your domain, so there is real value from the first session, and let your team's own skills accumulate on top over time. New hires then inherit both the starter judgment and everything the team has added since.
2. Keep the gate (this is the whole game)
The single rule that keeps a shared brain trustworthy is that a human approves every skill before it reaches the team. Drafts stay in the author's private workspace until then. This is the difference between Cognition and the wiki that rotted: there is no auto-indexing, so there is no garbage-in, and because there is no garbage-in, the team keeps trusting recall enough to actually use it.
Resist the urge to bulk-import your old wiki or doc dump on day one. Cognition is judgment, not an archive. Importing a wiki reintroduces exactly the landfill you are trying to escape: stale, unverified, unattributed entries that drag down recall for everyone. Capture skills as real work produces them, so every entry is something that actually worked, recently, for someone with their name on it.
3. Set up keys and scope correctly
Get the access model right early; it is annoying to untangle later. The model is simple:
- Org keys (cog_live_*) share the team brain; personal keys (cog_me_*) stay solo. Put the team on the org key so contributions land in the shared brain.
- Skills are org-bounded by default; nothing crosses organizations without explicit consent, so a partner org never sees your internal judgment.
- Keys are revocable at any time, and secrets and PII are redacted before anything is saved, so offboarding and safety are both handled.
4. Aim effort with team signals
Do not guess where the team needs help; the brain already knows and will tell you. Use these to point skill-capture and training at real gaps instead of assumptions:
"what is the team weakest on?" → get_team_weak_topics "how is our team retention?" → get_team_retention_summary "what has the team learned lately?" → get_team_recently_learned
Weak topics tell you where to capture new skills. Retention tells you what is decaying and needs a refresh before it misleads someone. Recently-learned is a ready-made digest for a standup or a weekly note, so the team sees the knowledge accumulating in real time.
5. Make the value visible, early
Adoption follows trust, and trust follows evidence. Run the stats and the activity ledger in a team setting once, where people can see hours saved and exactly which skills earned them. When the value is concrete and attributable rather than abstract, using Cognition stops being a mandate someone has to enforce and becomes something teammates ask for because they have seen it pay off.
A light cadence keeps it healthy without becoming a chore: capture skills as work happens, glance at weak topics and retention every week or two, and refresh anything flagged stale before it bites someone. That is the entire maintenance burden, and it is what keeps the brain sharp instead of letting it drift back toward the landfill.
next steps