Betsy Tong Warns: AI Amplifies Existing Problems, Not Solves Them

In a recent LinkedIn post, Betsy Tong offers a critical perspective on the current rush to adopt Artificial Intelligence (AI) within organizations. Rather than presenting AI as a magic bullet, Tong argues that it often serves to magnify pre-existing inefficiencies and data issues, leading to “confidently wrong results at 10x speed.” This journalistic coverage examines Tong’s insights into the foundational work required before implementing AI effectively.

The Perils of Scaling Messes with AI

Tong begins by challenging the common approach where organizations, upon forming an AI steering committee, immediately focus on purchasing new tools. She contrasts this with the reality of internal operational challenges that often go unaddressed.

The data cleanup needed for cloud migration is 5 years behind.
The vanilla ERP implementation created 19 more human workarounds.

According to Betsy Tong, this backwards order is a fundamental flaw. “AI amplifies what exists,” she states, emphasizing that without addressing underlying data quality and process issues, AI’s primary effect is to accelerate existing problems. Tong illustrates this with examples of AI-driven decisions that led to negative outcomes, such as biased recruitment screening and increased rejection rates, noting that in such cases, accountability is often diffused.

The Foundation for Successful AI Implementation

Betsy Tong asserts that the key to successful AI adoption lies not in the tools themselves, but in the groundwork laid beforehand. She highlights that initiatives which gain widespread adoption and deliver lasting value typically stem from addressing fundamental operational issues.

Prioritizing Data, Process, and Ownership

Tong outlines a three-step approach that should precede any AI tool implementation:

1. Audit Critical Data Sources: Identify the single most important data source for key decisions and audit it thoroughly before automating any processes that rely on it.
2. Simplify Core Processes: Find and simplify processes with the fewest edge cases. Tong argues that automation cannot fix a fundamentally flawed process; it only makes it faster.
3. Establish Clear Accountability: Assign a specific individual responsibility for overseeing AI outputs and empower them to halt the machine if necessary.

Tong stresses that this sequence—data, then process, then ownership—is crucial. As she puts it:

Before you touch the tools, do three things in order.
1️⃣ Pick the one data source your most important decisions run on.
Audit that before you automate anything that touches it.
2️⃣ Find the process with the fewest edge cases.
Simplify it first. Automation makes a bad process faster, not better.
3️⃣ Put a name on it.
Every AI output that drives a decision needs one person who can stop the machine.

In Betsy Tong’s view, the focus on flashy AI pilots often distracts from the essential, albeit less glamorous, work of cleaning data, redesigning workflows, and clarifying accountability. She concludes by urging organizations to remember that the problem may not be the AI itself, but the lack of a solid foundation upon which it is being built.