AI Decision Lab

Most AI training teaches people what to do with AI. It rarely helps them understand how their decisions shape AI outputs. AI Decision Lab was designed to make those decisions visible through a simulation-based experience focused on trade-offs, context, judgment, and the hidden reasoning behind effective AI-assisted work.

Overview

Status: Active Pilot

Client: Goldfin Consulting Inc. (internal)

Audience: Knowledge workers and cross-functional professionals integrating AI into operational workflows

My Responsibilities: Business Problem Diagnosis, Needs Analysis, Learning Experience Design, Pilot Design and Facilitation, Evaluation and Future Improvement

Tool Used: Facilitated via Microsoft Teams (breakout rooms), shared GPT environment

The Problem

Organizations are rapidly introducing AI into everyday work, but adoption remains inconsistent. Employees attend workshops, learn prompting techniques, and understand general best practices, yet many still struggle to apply AI effectively in real operational situations.

The challenge is not simply learning how to use AI. Real work involves ambiguity, competing priorities, incomplete context, stakeholder expectations, and time pressure. What works in a demonstration often breaks down when those conditions become real.

As a result, many people know the mechanics of using AI without developing the judgment needed to work effectively alongside it.

What The Experience Needed To Do

AI Decision Lab was designed to create practice around the judgment behind effective AI-assisted work, not just instruction on how to use AI tools.

The broader challenges around AI adoption, workflow integration, capability development, and human judgment extend far beyond a single learning experience. I explore some of those ideas further in my writing on AI enablement and organizational adaptation.

Related Thinking

Why AI Adoption Is a Workflow Problem, Not Just a Training Problem
AI Is Changing How Expertise Develops
The Hidden Risk of AI Adoption: Experience Debt
Why Human Judgment Still Matters in AI-Assisted Work

Strategic Design Decisions

Prompts are fixed by design. Participants cannot modify prompts. This intentionally removes prompting skill as the primary variable so the experience can focus on operational reasoning and decision-making instead.

One scenario, different priorities. All groups work from the same scenario and source materials, but each group operates under a different stakeholder priority.This design reinforces that effective AI-assisted work is context-dependent. The goal is not to produce a universally “best” output, but to make decisions appropriate for a specific situation.

Two input materials serve different purposes. Participants receive both a full meeting transcript and rough personal notes. The transcript contains everything that was said. The notes contain interpretation, prioritization, and implied context shaped by human judgment. This creates one of the central learning tensions in the experience: AI can only work with what it is explicitly given, while humans naturally infer meaning and importance.

Shared AI environment. All groups use the same shared GPT environment. This controls for differences in memory, personalization, and tool behavior so participant decisions remain the primary source of variation in outcomes.

Scoring focuses on alignment. Scoring remains hidden during the activity and is revealed only during the debrief.