A poorly written request costs more
A survey of more than 1,500 studies shows how the same request, worded two different ways, changes both the quality of an AI's answer and the cost.
Adriano Fante
Skills IT Blog
Technology for business: IT, security and AI in practice
All articles
Context matters more than the question
Stanford researchers found that AI makes more mistakes when the right answer sits in the middle of a long text. The problem is the material, not the question.
Written by SkyllerIA
The Long Conversation That Loses Its Thread
Microsoft and Salesforce research measured a 39% drop in AI accuracy across long, multi-step conversations. Here is why, and what to do about it.
Adriano Fante
Who sits on the AI committee — and what it checks
International research shows most AI committees can't prove what they verified after approving a policy. Here's the mechanism that changes that.
Written by SkyllerIA
The Right Question, the Wrong Document
Research from Stanford and Gartner shows the right AI answer depends more on the right document than on the model's raw intelligence itself.
Written by SkyllerIA
The Extension That Sees Everything You See
AI browser extensions ask to read and change everything you see on screen — including internal company systems. Here's what actually fixes that.
Equipe Skills IT
Request Templates for the Whole Team
A study of 1.4 million AI requests at work found that only 5% of employees use the tool well. Here is how to spread that skill to the rest of the team.
Adriano Fante
A Meeting Summary Is Not a Decision
Studies show most meetings never produce a recorded decision. Automated summaries help, but they do not replace an owner, a deadline, and a history.
Equipe Skills IT