Big Tech’s AI Bets, Platform Showdowns, and the Benchmarks That Actually Matter

In this edition: big tech strategy, platform power, and a surprising amount of reality-checking.

This is Q&AI, our blog series aimed at keeping you in the know on updates in the rapidly evolving world of AI. Sometimes, these will be quick updates on new developments in the field. Sometimes, they’ll be tips on tactics, features, or functionality. 

If you haven’t met me yet, hi: I’m Jen Taylor, CI’s Director of AI Strategy & Implementation, and your (very human) AI BFF. AI is moving at the speed of light, so I’m here to let you know what matters most now. 

Q: What’s happening with Siri and generative AI? 

A: Siri has been far behind its competitors when it comes to generative AI. For months, the industry has wondered whether Apple would build its own foundation model or borrow someone else’s. Now we know: Apple is partnering with Google to use Gemini as the backbone of a custom Siri rebuild.

This decision reportedly costs Apple $1B per year and is intended to be temporary — a stopgap until Apple can develop its own internal model. It’s a fascinating moment: Apple, famous for building everything in-house, is renting intelligence from a rival to stay in the game.

Q: Is Google really exploring AI data centers in space?

A: Yes — and it’s as wild as it sounds. Project Suncatcher is Google’s new moonshot initiative to explore whether AI training could eventually take place in space, powered directly by the sun.

The potential upside? Near-infinite clean energy, and a way to ease the enormous environmental footprint of AI.

The challenges? Keeping satellites in ultra-tight formation, building an orbital compute infrastructure, and moving data between Earth and orbit efficiently.

Moonshot is the right word…but if it works, it could fundamentally reshape the energy equation of large-scale AI.

Q: What’s going on with Amazon’s cease-and-desist letter?

A: Amazon has accused Perplexity’s shopping agent of covertly accessing Amazon customer accounts and disguising automated activity as human browsing. This dispute is an early preview of what’s coming as agentic AI tools begin navigating the web on our behalf:

• Will platforms block third-party AI agents?
• Is this about protecting people or protecting market share?
• How much autonomy should browsing agents have?

This is likely the first of many conflicts between incumbent platforms and AI agents that bypass the traditional user interface.

Q: What did this year’s McKinsey AI survey reveal?

A: McKinsey surveyed 2,000 organizations, and the results show a clear pattern: 88% of companies now use AI in at least one business function…but only 33% have scaled AI beyond pilots.

The key takeaway: AI creates value when it redesigns workflows, scales across teams, and drives growth, not when it’s limited to small experiments. In other words–the hype is widespread, but the hard work of operational change is just beginning.

Q: Do current AI benchmarks actually measure whether AI can do work?

A: Most benchmarks test intelligence, not capability. That’s the gap the new Remote Labor Index aims to close. It measures how autonomously LLMs can actually perform real work, rather than how smart they appear.

Good news for humans: automation rates across the leading LLMs are still extremely low. The highest score so far is 2.5% full autonomy. This feels like a more grounded way to measure AI progress — and a promising sign that we’re nowhere near “AI can do everyone’s job” territory.

Your Friend,
Jen

Have a question for a future edition? Submit it here!