How Much User Choice Remains When AI Features Become Default Options?

Recently, I've been looking into the UK Competition and Markets Authority (CMA)'s investigation into Microsoft, triggered by the price hike after introducing Copilot to Microsoft 365, where users may not have been informed of alternative options. As a researcher who just joined an AI Lab, this news made me ponder a recurring question: Is the commercialization of AI capabilities 'value-add' or 'bundling'?

First, let's look at the data. In January 2024, Microsoft raised the prices of Microsoft 365 Personal and Family subscriptions by 30% respectively (Personal from $69.99 to $99.99/year, Family from $99.99 to $139.99/year). Simultaneously, Copilot was integrated directly into the subscription, with no cheaper alternative offered that excluded Copilot. In other words, if users only want basic Office functions, they must pay extra for AI features.

The UK CMA's concerns are specific: Did Microsoft fail to "clearly explain optional alternatives" when adjusting subscriptions? This points to a deeper issue—when AI capabilities become default components of a platform, can users still make choices based on "informed consent"?

I outlined several key elements of the incident:

• Pricing Changes: Personal annual increase of 30%, Family increase of 40%, both exceeding general inflation rates.
• Feature Bundling: Copilot is defined as a "value-added feature," but users cannot choose not to pay for it.
• User Base: Microsoft 365 has over 700 million global subscribers, with personal versions accounting for a significant proportion.
• Regulatory Impact: After the CMA investigation launched, Microsoft offered UK users an "older subscription version without Copilot," but only for new users; existing users still need to manually switch.

Why is this worth AI researchers' attention? Because we are witnessing the transition of AI products from "technical demos" to "business models." Copilot requires substantial computing power, and Microsoft's pricing strategy is essentially spreading the cost of AI infrastructure. The question is whether this cost-sharing is fair.

From a paper perspective, a recent article titled The Economics of AI Bundling mentions that bundling sales in AI products are more dangerous than in traditional software, because the marginal cost of AI features is high (inference costs), and user demand for AI varies greatly. Some users use Copilot dozens of times a day, while others might not use it once a year. Uniform pricing leads to cross-subsidization—low-frequency users paying for high-frequency users.

Another point worth discussing is "cognitive load." Microsoft emphasized "Copilot value" in its price hike notices, but users need to understand what AI features actually do. According to Microsoft's Q3 2024 earnings report, enterprise adoption of Copilot for Microsoft 365 reached 3.8% (among enterprise users), but personal version data was not disclosed. This suggests that ordinary users' actual usage rate of AI features may be far lower than expected.

A table comparison clarifies the issue:

Dimension	Old Plan	New Plan (with Copilot)
Annual Fee (Personal)	$69.99	$99.99
Core Features	Office Suite + 1TB Cloud Storage	Same + Copilot
User Choice	Optional	Mandatory Bundle
Inference Cost Sharing	None	All users bear it
UK Regulatory Status	Fully Compliant	Under Investigation

The UK CMA's intervention is actually testing a boundary: When AI features become "standard" on a platform, does it constitute monopolistic behavior? Under Section 46 of the UK Enterprise Act 2002, if a company misleads consumers regarding "optional alternatives," it can be fined up to 10% of its turnover. Microsoft's global revenue exceeds $200 billion, so this figure is not negligible.

But the deeper question is: Should the commercialization of AI technology follow the principle of "user-driven" choice? My view is that AI features should exist like "plugins" rather than "glue." Users should be able to choose whether to enable them, whether to pay for them, rather than being forced to upgrade. This aligns with consumer rights and benefits healthy iteration of AI technology—because only features genuinely needed by users receive real feedback and improvement.

Data indicates that mandatory bundling may actually reduce willingness to use AI. A survey by Consumer Reports showed that 62% of users stated they would consider abandoning the Office suite for free alternatives (like Google Docs, LibreOffice) if AI features caused price hikes. This suggests that the pace of AI commercialization needs more refined design.

Returning to a research perspective, I believe this case serves as a warning to colleagues in AI Labs: Do not package AI capabilities as "default services" without user knowledge. Transparency and choice are the foundations of trust, and trust is the prerequisite for the long-term commercialization of AI products.

To summarize the core viewpoint in one sentence: When AI features are forcibly bundled, users lose not only their right to choose but also their understanding and trust in the essence of "intelligent services."

Original Link: https://www.ithome.com/0/983/268.htm