How AI Helps the Best and Hurts the Rest

Why It Matters

The findings show that generative AI can widen performance gaps, helping the already successful while hurting weaker firms, underscoring the need for judgment‑focused deployment and safeguards.

Key Takeaways

• Top‑half firms gained ~15% revenue and profit with AI advice
• Bottom‑half firms lost ~10% revenue and profit using same AI
• Both groups asked similar questions and received comparable suggestions
• Success depended on entrepreneurs’ ability to filter generic advice
• Leaders must embed judgment checks to prevent AI‑driven inequality

Pulse Analysis

The Kenya study illustrates a broader truth about generative AI: its value hinges on the user’s capacity to interpret open‑ended recommendations. In narrow tasks—drafting emails or coding snippets—AI can lift even low‑skill users. But when the problem space is vague, such as setting pricing strategy or choosing new product lines, the tool merely surfaces possibilities. Entrepreneurs with strong business acumen can extract context‑specific insights, like adding gaming accessories to a cybercafe, and translate them into profit‑driving actions. Those lacking that judgment tend to adopt generic, margin‑eroding advice, such as indiscriminate price cuts, leading to poorer outcomes.

For managers and policymakers, the experiment warns against blanket AI rollouts. Deployments that focus solely on average effects may mask harmful impacts on vulnerable groups. Effective implementation should therefore include layered guardrails: richer user context fed into the model, decision‑checklists, and escalation pathways for high‑risk choices. Training programs that sharpen questioning skills and critical evaluation can amplify the benefits for lower‑performing firms, turning AI from a divider into an equalizer.

The broader implication for the global economy is clear. As AI tools become cheaper and more ubiquitous, firms must treat them as decision‑augmentation rather than decision‑replacement technologies. By measuring heterogeneous effects, investing in human judgment, and designing context‑aware interfaces, organizations can harness AI’s scalability without amplifying existing inequalities. This nuanced approach will determine whether generative AI truly democratizes business insight or simply accelerates the success of the already advantaged.