Artificial intelligence can help firms choose actions that raise demand or reduce costs, such as which product variant to launch, which price or promotion to offer, how to target advertising, how much inventory to hold, or when to expand capacity. But AI may also make competitors' decisions more alike. We study this trade-off in an oligopoly model where firms choose whether to adopt an AI decision-support technology before taking state-contingent actions. The AI system may provide either a forecast of market conditions or a direct recommendation about which action to take. Adoption improves the accuracy of this information or advice. When multiple firms adopt related AI systems, their recommendations may also become correlated, so that competitors are more likely to be right or wrong together.

The value of adoption depends on the strategic payoff to being correct. If a firm benefits most from being correct when its rival is wrong, AI adoption decisions are strategic substitutes: rival adoption reduces the incentive to adopt, and more correlated AI advice makes joint adoption less attractive. If instead a firm benefits most from being correct when its rival is also correct, adoption decisions are strategic complements: rival adoption increases the incentive to adopt, and correlation reinforces joint adoption. The same AI decision-support technology can therefore lead to asymmetric adoption, full adoption, no adoption, or multiple equilibria depending on the nature of downstream competition.

We then study an upstream AI supplier that chooses how to price and design the technology. The supplier's preferred business model depends on the same strategic forces. In markets where firms value differentiated correctness, the supplier may prefer selective access or less correlated recommendations. In markets where firms value coordinated correctness, it instead prefers broad access and standardized recommendations. The analysis shows that the competitive effects of AI depend not only on the accuracy of its advice, but also on whether AI preserves or reduces informational diversity among rivals.