Jeff's Daily Dose: why Meta won & Microsoft lost

Yesterday delivered the perfect case study in AI investment.

ICYMI: Meta's stock jumped 10%. Microsoft's cratered 10%... wiping out $357 billion in a single day.

Same week. Same AI spending story. Opposite results.

The difference? Meta proved that its AI actually does something. Revenue up 22% year-over-year. Ad clicks up 3.5%. Real conversions, real money.

Microsoft? A record $37.5 billion in capital spending. Azure growth slowing. Investors finally asked the uncomfortable question: "Where's the payoff?"

This is becoming a common pattern: The AI pilot doesn't translate to profit.

But here's the thing... you're not Microsoft.

You're far smarter & more disciplined. You don't have billions to burn on infrastructure hoping that something sticks. Which means you can't afford their mistakes.

So, here's how to be sure your AI investment actually pays off:

(1) Tie every tool to a measurable business outcome before you buy it. Meta didn't just deploy AI... they deployed it specifically to improve Ad Conversions. If you can't name what number moves, don't spend the money.

(2) Start with your highest-volume repeatable task. Meta focused AI on their ad-ranking model... something they do billions of times daily. Find your equivalent. Is it Recruiting? Onboarding? Customer response? Service tickets? Pick one thing you do constantly & automate that first.

(3) Upskill two people deep, not one. Meta's CFO didn't just report AI results... she understood the architecture changes that drove them. Your AI champions need backups. When one person holds all the knowledge, you've built fragility, not capability.

(4) Set a 90-day ROI checkpoint. Microsoft kept spending without clear feedback loops. You won't. Before deploying any AI tool, define what "working" looks like and when you'll measure it.

(5) Kill what isn't working. This is the hard one. Microsoft kept doubling down. Meta doubled their GPU usage on what was working (ad ranking)...not everything. Be ruthless about stopping pilots that don't perform.

The "people part" is the whole game here.

AI tools are cheap. Training is available everywhere. The companies that win are the ones whose people actually use the tools... consistently, skillfully, in ways that hit real metrics.

That's a leadership problem. Not a technology problem.

How are you measuring AI payoff in your organization? Share with the Roundtable, so we can learn together.

4 comments