Coordinate and act
Verification before dissemination
Check before you share, hold the fast-moving item for one more look, and confirm consequential claims against a second independent source.
Borrowed from newsroom practice, where reporters confirm a story before publishing it, even if a competitor publishes first. Read the original source.
Ongoing habit · Minutes to start · Anyone on the team can start it
Try this first
One rule. Nothing consequential ships unverified, and anything moving fast gets one more look.
Use it when
Speed is being rewarded over accuracy.
Skip it when
Speed matters sometimes and not everything warrants a hold. Spend the discipline where being wrong is costly.
How to introduce it
Verify before you share or act, not after. When something is moving fast and looks exciting, hold it for one more check. Newsrooms learned this: people forget who was first and remember who was wrong. Apply it to trusted colleagues and confident models alike. For a consequential claim, the check means a second independent source. One source, human or AI, is a lead, not a fact.
How to show up
Reward getting it right over getting it out, visibly, so the incentive matches the rule. When something is racing toward release, be the person who asks for one more check.
How long it takes
No setup. It's a standing discipline.
What makes it hard
A fast, impressive result pushes hard against holding it, and that pull peaks exactly when the check matters. If the team rewards speed over accuracy, the rule will not survive, so fix the incentive first.
What it looks like when it's working
Consequential work gets checked before it goes out and the team makes fewer public errors. A reflex to move fast on confident AI output means the practice is not changing behavior.
How long until it sticks
A few weeks, and ideally a caught error or two, before verify-first becomes the reflex.
How you know it stuck
People verify before sharing and hold the exciting item without being told.
The idea behind it
Right beats fast. The memory that lasts is who got it wrong. A fluent AI answer creates the strongest pull to skip the check.
Where it comes from
Newsroom practice. Reward getting it right over getting it out. Make "who got it wrong" the memory, not "who was fastest." The core rule is to check before publishing, and to hold a hot item for one more call, because people forget who got it first but remember who got it wrong.
Also fits: Confident claims go unchecked