Don't dump it all into a second brain

Dumping everything your business knows into a second brain is the quickest way I know to get confident, wrong answers from AI.

Second brains are all the rage, and everybody's rushing to fill one for their AI to work from. But AI treats everything in there as equally true.

Your shared drive probably has three versions of the same document: the current one, the old one, and a draft nobody approved. You know which one is current because you were there when it changed. AI has no way to know that, so it builds a confident answer on whichever version it finds.

I learned this on my own second brain. When I audited it, about half of nearly seven thousand records were no use to AI at all, and every one of them was a chance for it to pick the wrong file.

So now I start small, with the twenty files that matter most for one job. At the top of each one, add three labels: where it came from, who owns it, and the date it was last checked.

The date shows you and your AI whether the file is still current. The owner is the person who updates it when something changes, so it doesn't quietly drift out of date again.

Then give AI those twenty files and nothing else. You can add more once those twenty are giving you answers you'd trust in front of a client.

With a second brain, quality matters far more than quantity. Start with the twenty files for one job this week.