Microsoft is reportedly shifting away from exclusively using OpenAI models for its Microsoft 365 Copilot assistant, aiming to enhance efficiency and reduce costs. The company is developing smaller in-house artificial intelligence models and considering third-party options, as reported by Reuters. Microsoft has invested billions to secure exclusive rights to OpenAI’s GPT-4 model, which powers Copilot, priced at thirty dollars per user per month. The transition comes amidst a growing partnership with a new AI subsidiary led by Mustafa Suleyman, focusing on proprietary models. Microsoft CEO Satya Nadella closely monitors these developments, including integrating new AI models from Anthropic and Google into GitHub Copilot.
I want to make a point with this story: For most customers, if the core useful product is CoPilot, then they should not care. They do not care what model drives the product.
More broadly, as the vast majority of customers are likely getting AI integrated into products, they, too, would not care which model.
I use the wine analogy a lot when talking about AI. If the models are grapes, and you sell red wine, you generally only care about the quality of the wine. The choice of grapes may impact quality, and different grapes may have matches to different dishes, and the expertise to match to a quality meal is the value. With AI, you sell solutions, and the quality of the models is a component of the outcome, not the entire solution.