Lucidworks

@Lucidworks

AI search and discovery that connects people to the products, answers, and knowledge they need.

San Francisco, CA lucidworks.com Joined January 2009

2.1K Following 6.9K Followers 7.9K Posts

Latest posts 5

Step 1 of 6 to a trusted AI answer: Connect.

An agent that can reach one system gives partial answers with full confidence. Lucidworks® connects SharePoint, Confluence, ServiceNow, file shares, and more. na2.hubs.ly/H08yFLl0 #AIagents

Lucidworks @Lucidworks 6d

A right answer built on content the asker was never allowed to see is worse than a wrong one.

Permissions have to travel with the content and be checked on every question. na2.hubs.ly/H08yFMP0 #AIgovernance

Lucidworks @Lucidworks Oct 30

Bigger context windows make it tempting to give an agent everything. Then the passage that answers the question competes with hundreds that don't.

Give agents a small, ranked set of the right passages instead. na2.hubs.ly/H08yFMN0 #AIagents

Lucidworks @Lucidworks Oct 27

Three ways AI agents go wrong, and none of them is the model:

Too little context: it guesses.
Too much: it loses the thread.
The wrong context: it overshares.

The fix starts with what the agent is shown. na2.hubs.ly/H08yFMQ0 #AIagents

Lucidworks @Lucidworks Oct 23

AI agents don't fail on models. They fail on context.

Too little, and they guess. Too much, and they lose the thread. The wrong context, and they overshare.

Eight weeks on getting it right starts now: na2.hubs.ly/H08yFPQ0 #AIagents