"Without data, you're just another person with an opinion."

Reading time: 3 min. 

A new KPMG Global AI Pulse survey of 2,145 C-suite leaders across 20 markets finds only 7% have established ROI from AI. Yet 79% name it their top investment priority. The spread is not a technology problem. It is an accountability problem.

Where the CEO directly owns AI outcomes, established ROI reaches 14%. Where a committee does, it falls to 4%. Organizations with full AI cost visibility are 5x more likely to report established ROI.

Most organizations already have the budget. What separates the 7% is ownership: one name on the line, cost visibility before scaling, success defined in output terms before the pilot starts. My book, 100x, shows how.

To celebrate the second edition, I'm giving away 100 copies to the first 100 voters in the poll below 👇🏽

How human-centric AI delivers business results

Your weekly briefing on tools that create competitive leverage

Meta Superintelligence Labs released Muse Spark 1.1 this week with the most complete four-model comparison to date. Scores are directional: Meta ran competitors on its own scaffolding.

Muse Spark 1.1: your agents need to run professional workflows. Order-to-cash, S&OP, financial analysis, compliance. Leads white-collar task execution across 220 enterprise tools. The model for CIOs and Chief AI Officers putting agents into production workflows.

Claude Opus 4.8: your agents need to operate software or produce documents. Desktop automation, board decks, financial reports, software engineering. Strongest when agent output lands in a human's inbox.

GPT-5.5: your engineering team is building agents on large codebases. Long-horizon coding, terminal automation, million-token document retrieval. The foundation for teams building agent infrastructure, not running it.

Gemini 3.1 Pro: your stack is already Google Cloud. Trails on every enterprise benchmark (JobBench: 15.9, last of four). Outside Vertex AI and BigQuery consolidation, the data does not support it.

When her aunt got lost navigating dementia care for Jenny Lee's grandmother, she saw the real problem: families leaving hospital discharge with no guide and no budget for one. Her company Hera pairs AI-encoded local healthcare knowledge with credentialed care managers to fill that gap. This month, Hera raised $27 million Series A with referral partnerships at Weill Cornell, NYU Langone, and UCSF. Ninety-five percent of families stay enrolled.

The 7% who have established AI ROI assign accountability differently. They also changed the unit of time. Annual planning assumes this year's competitive distance roughly equals last year's. AI compute costs fall 10x annually. Organizations on yearly cycles are planning for a world that won't exist when the plan arrives.

What would it take to run your next AI initiative as a twelve-week cycle instead of a twelve-month plan?

👉🏽 Read the full framework in AI in cycles scales faster.

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

In Sprint, Jake Knapp (Google Ventures design partner) lays out a five-day process for validating critical decisions before committing budget. The logic maps directly to AI deployment: compress the validation window and let a real user decide what ships.

Perfect for: Leaders who need a repeatable method to prove AI value before scaling.

When AI accountability belongs to everyone, it belongs to no one.

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!