Personal portfolio of Beza Lefebo. The views here are my own. This site is not associated with any employer or client, and it does not offer or sell any services.
BezaSpeaksPersonal portfolio

Personal portfolio · Thought leadership

Notes on AI, trust, and how people really learn to use it.

I'm Beza Lefebo. This is my personal space for thinking out loud about artificial intelligence: why responsible AI has to be shown rather than claimed, why most AI awareness training leaves work unchanged, and what learning AI role by role could look like instead. Everything here is my own view, shared to start conversations, not to sell anything.

Read the blog Why AI training falls short
Beza Lefebo
Beza LefeboD.Eng., The George Washington University

What I think about

Five threads I keep pulling on

Each one is a question I research and write about. Start anywhere.

  1. 01 Responsible and trustworthy AI AI fails quietly. Trust has to be built in layers and shown with evidence, not claimed in a values statement. Read the essay
  2. 02 Customized, role-based AI training AI helps on one task and hurts on the next, so training has to be built task by task. That means role by role. See the argument
  3. 03 DRILL Do, Reflect, Improve, Learn, Launch. People learn AI by producing first and understanding why it worked afterward. How it works
  4. 04 BELGIC Build, Explore, Learn, Grow, Innovate, Control: six AI capabilities, weighted differently for every role. The six capabilities
  5. 05 Beza on AI 51 short essays on the AI race, governance, safety, business value, and what AI means for the people doing the work. Browse the blog

The idea behind most of my writing

“The organization has a completion rate. It does not have a capability.”

In a field experiment with 758 consultants, AI raised the quality of work on tasks inside its capability by about 40 percent, and cut correct answers by 19 percentage points on a task just outside it. Researchers call this the jagged frontier.

If AI's usefulness changes task by task, a single all-staff course cannot tell a budget analyst, a program manager, and an auditor where the line falls in their own work. That gap is what I study.

Dell'Acqua et al., 2023, Harvard Business School Working Paper 24-013.

Read the full argument

Beza on AI

Recent essays

All 51 essays
The global AI raceKimi K3 and the End of American AI ComplacencyFor several years, the American AI industry operated under a comfortable assumption: China might produce capable models, but the most advanced systems…2 min read Safety and securityWhen Guardrails Block the DefendersThe recent Hugging Face security incident exposed an uncomfortable contradiction in the way the AI industry discusses safety.2 min read Safety and securityThe OpenAI Agent Incident: When AI Escapes the AssignmentFor years, conversations about AI risk were dominated by distant scenarios: superintelligence, machines taking control, or systems developing intentions…2 min read