Meta signals AI-driven workforce reduction
The hosts opened with breaking news about Meta’s reported plans to reduce its workforce by up to 20%, from 80,000 to 64,000 employees. According to Reuters, this would increase revenue per employee from $2.2 million to $3.5 million and could drive 3% to 5% earnings per share upside in 2026 and 4% to 7% in 2027.
Brian noted that when he asked attendees at the recent Master B2B Summit whether AI would reduce their headcount, every hand went up, even though the panel discussion itself offered the public line that AI would not cut jobs. The disconnect between public messaging and private expectation is striking.
The scale of investment is substantial. Meta reportedly plans to spend $135 billion on AI infrastructure this year and $600 billion by 2028. Google is at $180 billion and Amazon at $200 billion. Andy observed that companies cannot spend heavily on both people and infrastructure, so Meta appears to be choosing infrastructure.
I remember the days when companies would announce layoffs and their stock price would go down. Layoffs, stock price goes up. This is crazy.
Brian Beck, Master B2B
Apple’s alternative approach: charging tolls instead of building
The hosts contrasted Meta’s approach with Apple’s. Apple is reportedly spending relatively little on AI infrastructure (in the low double-digit billions) but has generated $1 billion in AI revenue by charging other companies for access to its ecosystem. An estimated 80% of that revenue comes from ChatGPT subscription fees that flow through Apple’s platform.
The hosts described this as a toll model. Apple does not need to build AI capability if it can charge others for access to its user base. Brian compared it to a speakeasy: good to be Apple when you control the gate.
The paradox of freed time
The main topic turned to whether AI is creating production or productivity. A Bloomberg article cited a claim that 40% of worker time is freed up by AI, but the hosts questioned whether that matches reality.
Andy cited ActiveTrack research showing that the share of workers at risk of disengagement (defined as underutilized at least 75% of the time) rose from 19% after ChatGPT’s release to 23%. Some workers are using AI to do their homework, so to speak, and then disengaging from productive work.
At the other end of the spectrum, a Wall Street Journal article argued that AI is not lightening workloads but making them more intense. The capacity that AI frees up immediately gets repurposed into additional work. The phrase cited was that the appetite grows with the eating.
After AI adoption, every work category increased
The hosts shared data from a 2026 State of Workplace Report. After AI adoption, every single measured work category increased: email up 104%, messaging up 145%. Nothing decreased. The promise of doing less with AI is not showing up in the data.
Only 3% of workers are in what the report called the productivity sweet spot, defined as spending 7% to 10% of their time in AI tools. Meanwhile, 57% spend less than 1% of their time there. Andy compared this to racing: if you put the pedal to the metal the whole time, you burn out. If you never accelerate, you lose. The strategic and tactical use of AI at the right moments produces disproportionate gains.
The paradox of choice and information overload
Brian drew a parallel to Paco Underhill’s book Why We Buy, which argued that overwhelming customers with too much choice causes them to buy nothing. The same dynamic may apply to AI-generated information: the tools prompt you to go deeper and deeper, and before you know it, two hours have passed.
Andy cited the paradox of choice research: if you give people a hundred choices, they will choose from within two or three. Offering more choices causes them to snap in the other direction. The abundance of information that AI can generate may not be helpful if it overwhelms rather than clarifies.
Production versus productivity: a summit conversation
The hosts recounted a discussion from their recent summit. A marketing communications professional said she could now produce many more press releases with AI. Andy asked whether that was productivity. She said yes. He asked whether the quality was the same as what she would produce herself. She said no.
This exchange crystallized the distinction. Productivity should mean doing the same quality work in less time. If you are doing more work of lower quality in the same time, that is production, not productivity. The question becomes whether quality matters for a given task.
You’re producing more, but are you more productive? Because if you have to go back and fix all that stuff or it lowers the standard, then it begs the question: what is productivity?
Andy Hoar, Master B2B
Community poll: doing more work wins
When polled, 67% of the Master B2B community said AI is causing them to do more work, while only 33% said it lightens their workload. Both hosts acknowledged experiencing this themselves. The freed time gets filled with additional tasks rather than used for rest or personal time.
The challenge is deciding when to use AI to drive work versus when to use it to enhance work. Right now, AI is capable of enhancing but may not be capable of driving. Some people are using it to drive and are getting ahead of the technology. Others are not using it at all and are falling behind. The threshold will shift as the technology improves, but for now, the strategic users are those in the 7% to 10% range.