Hello, dear readers!

Back in Issue #92 we looked at how GLP-1 drugs were quietly rewriting America's shopping cart. This time we're going behind the counter, to the company that turned all that appetite suppression into the most valuable healthcare business on the planet. Eli Lilly is worth a trillion dollars now, and the way it got there is a masterclass in speed, pricing, and customer ownership.

Also, one startup founder put it perfectly: AI is doing to company headcount what Ozempic does to waistlines. So why are some of America's biggest employers suddenly hiring again?

This issue features the usual bunch of AI Insights and recommendations for the week's podcasts, books, shows, charts, and tweets, followed by a final chuckle.

Let's get going.

Industry Highlights: How Eli Lilly Got Huge by Making Us Small

When we covered GLP-1s in Issue #92, the lens was on us, the shoppers eating less, drinking less, and skipping the snack aisle. The bigger story was always going to be who cashed in. The answer, as it turns out, is a 150-year-old drugmaker from Indianapolis.

A trillion-dollar waistline

Eli Lilly is now worth about a trillion dollars, one of only a dozen or so companies to cross that line, and almost all the others are in tech. Its diabetes shot Mounjaro has passed Merck's cancer blockbuster Keytruda as the world's bestselling medication. Its obesity shot, Zepbound, is the planet's most popular weight-loss drug, barely two years after launch. Together, they now account for roughly half of Lilly's revenue, and the stock is up nearly 300% since Mounjaro won approval in 2022. CEO Dave Ricks says the company ships seven Zepbound injections every second.

Here's the part worth studying. Novo Nordisk had the first blockbuster GLP-1s in Ozempic and Wegovy, and Lilly still ended up on top. Why? Execution.

Speed, vials, and owning the checkout

Ricks, CEO since 2017, rebuilt Lilly around speed and a low tolerance for mistakes, halving the usual drug-development timeline and running the research group like a startup. When demand exploded and both companies ran short, Novo was stuck rationing doses because of an ingredient bottleneck. Lilly's bottleneck was its complicated injector pens, so it simply switched to cheap single-use vials, then cut the vial price in half to $399 a month through its own direct-to-consumer site, LillyDirect. Today, 55% of new Zepbound patients buy that way. It also priced Zepbound about 20% below Wegovy and hired Nike's ad agency, Wieden and Kennedy, to sell Lilly as a friendly "medicine company." As one BMO analyst put it, once you make it hard for customers to get your product, you permanently change their behavior. Novo assumed its early lead would hold, and that assumption is proving expensive.

Read that sequence again, because it's a procurement story wearing a pharma costume. Both companies hit a constraint at roughly the same moment. Novo protected its format and rationed. Lilly looked at its own bottleneck, decided the fancy delivery device was not the product, and shipped in cheap vials while its competitor was explaining shortages to patients. Every one of us has sat in a meeting where somebody defends the spec rather than the customer, usually with a very good reason attached. This is what that costs.

The victory has a shelf life

None of this makes Lilly bulletproof, and Ricks knows it. Every blockbuster meets the patent cliff. Lilly has roughly a decade of protection left, but at least 120 companies, a third of them in China, are chasing the same category, and prices are falling faster than investors expected. Knockoff peptides are already circulating on the grey market under names like Reta, Triple G, and even “ratatouille.” Then there's the newer threat: AI drug discovery, which is why Lilly built its own Nvidia supercomputer, a billion-dollar San Francisco lab, and a data moat it describes as three million failed drugs' worth of in-house research. Ricks is hedging with a $20 billion acquisition spree this year, including a $7.8 billion bet on sleep disorders that he thinks could be the next obesity.

The bottom line for business readers is an old one dressed in new clothes. In a gold rush, the miner who controls supply and owns the customer relationship beats the one who struck first. But no moat outlasts the patent clock, which is why the smartest thing Lilly is doing right now is spending like a company that knows this party ends.

The Future of Work: GLP-1 for the Org Chart

Jay Reno's company, Pointhound, helps people book flights using credit card points. Last year 750,000 people used its products. It has four full-time human employees and a fleet of AI agents. Reno's previous company, a furniture-rental business he sold in 2022, had around 150 people at its peak. His description of the difference? Building with AI is “like everyone was just given the ability to take a GLP-1.”

That metaphor is doing a lot of work, and the data backs it up. A new working paper by Rem Koning at Harvard and Hyunjin Kim at INSEAD examined thousands of Y Combinator-backed and other venture-backed startups. The AI-native ones run with about 25% fewer employees, 15% fewer entry-level workers, and 15% fewer managers, with fewer rungs on the ladder, yet they raise similar money and carry similar valuations. Startups selling services like tutoring or mental health were up to 70% smaller than their non-AI peers.

The giants are copying the homework. When Microsoft's Xbox division cut 3,200 people this month, its CEO, Asha Sharma, said the plan was to compress management from as many as 14 layers to three to five. “We will simplify,” she wrote. Amazon and Meta are de-layering on the same logic. At the AI-powered investment bank OffDeal, engineers and bankers sit side by side at nearly a one-to-one ratio; the firm had engineers shadow bankers for two weeks to automate the prep work, and skipped hiring junior analysts on purpose, so the process would have to get smarter.

So is this just a smaller world with fewer jobs? Not so fast. After a year of AI-driven hiring freezes, big employers are quietly staffing up again. Alphabet, CSX, Booz Allen, Snap-on, and ServiceNow have all told investors they need more people, and US jobless claims recently hit their lowest since 1969. Lattice CEO Sarah Franklin says the companies that gutted their entry-level ranks, assuming agents could cover the work, have realized humans still have to run alongside the machines: “Just because you have coding agents doesn't mean you're not hiring engineers.” MIT's Paul Osterman is blunter about where all this lands. Do companies need more people or fewer? “We have no idea. No one has any idea.”

The headcount numbers grab the headlines, but the bigger change is structural: flatter reporting lines with AI wired straight into the workflow, which Koning notes is the part no tool delivers on its own. The real trap is over-shrinking the junior pipeline, because that's where tomorrow's senior talent and the freshest feel for these tools actually come from.

What I'm seeing in supply management is less hand-wringing about the entry-level pipeline and more focus on practical use cases that actually move productivity. And here's the part that surprised me: companies aren't holding onto their junior hires out of sentiment. They're expecting those hires to be the AI-native ones, the people who get the tools adopted across a team that has been doing it the old way for fifteen years. Turns out the junior analyst isn't competing with the agent. They're helping drive others on the team to use it as well. 

AI Insights

Your shared chats were showing up on Google (BBC): Hundreds of shared Claude conversations turned up in ordinary search results, some exposing users' CVs, contact details, and proprietary work research they'd pasted in. Anthropic says shared links were never meant to be discoverable and has pulled them from search. Still, "share" and "publish" sit closer together than most people assume, and that's worth a reminder to anyone on your team pasting contract language into a chatbot. 

The people building it want a brake pedal (Yahoo): More than 1,000 staff at frontier labs, including OpenAI's chief scientist and several Anthropic cofounders, signed an open letter pressing Washington to back tools that can deliberately slow the pace of automated AI research. It landed right after OpenAI admitted that two test models slipped out of its sandbox, reached the open internet, and hacked another company's systems. Even Sam Altman now says it may be time to tap the brakes. 

The grid pushes back (TechCrunch): PJM, the biggest US grid operator, will start cutting power to data centers of 50 megawatts or more during shortages beginning June 2027, after a capacity auction fell short. Demand is set to quadruple by 2035, so expect a rush toward on-site generation and dirtier diesel backups. Of course, this is the pattern we keep seeing: the AI boom is quietly becoming an electricity story, and electricity is a supply chain like any other.

The Supply Aside

📕 Read: The Age of Turbulence: Adventures in a New World by Alan Greenspan

Alan Greenspan died in June at the age of 100, and the man they called “the Maestro” ran the Federal Reserve longer, and with more mystique, than almost anyone alive. This 2007 memoir, written after eighteen years in the command room of the global economy, is his own reckoning with booms, busts, and the limits of anyone's ability to see them coming. Fitting reading in a week when everyone is arguing about whether the next boom, obesity drugs or AI, take your pick, is a miracle or a bubble. The man who coined “irrational exuberance” opens the book by cheerfully admitting it probably contains errors he cannot find.

What Else I’m Reading

Don't blame supermarkets for your expensive groceries (The Economist): Everyone wants a villain for the $9 pack of ground beef, and the supermarket is the easy target. This is something The Economist calls a populist myth. Grocer margins are razor-thin, roughly 1% at Kroger and 4% at Walmart, while the packaged-food giants run closer to 13%. The real culprits are fertilizer, fuel, and the war in Ukraine, not the checkout.

Big companies are starting to hire again (The Wall Street Journal): After the AI-wipeout predictions, the reversal is quiet but real. Booz Allen, which cut 7.5% of its headcount last year as federal contracts dried up, now says it needs to speed up hiring for security-cleared national-security work. Robert Half's CEO reckons AI's hit to the job market is proving "more benign than some have feared." The staffing data is starting to agree with him.

The art of knowing when to step down (The Financial Times): Pilita Clark on one of the hardest tricks in any career, knowing when to walk away. She lines up the leaders who cling on, from Sumner Redstone to a bench of octogenarian US senators, against those who plan the handover, like FedEx's Fred Smith grooming his successor and Warren Buffett bowing out of Berkshire at 95. Planning your own exit, she argues, is part of the job.

📺 Watch: The Economist's The Insider: An Interview with Elon Musk

For his first extended interview since SpaceX's blockbuster IPO, Musk spends 85 minutes at his Texas Gigafactory with Zanny Minton Beddoes, The Economist's editor-in-chief. They cover his timeline for AI overtaking human intelligence, China's pace, his interest in Europe, and why he now thinks he “got a little too involved” in politics. Worth the time for how a man worth nearly a trillion dollars talks about the future he's betting the company on.

👂 Listen: The Knowledge Project: “The Mindset That Unlocks Your Full Potential” with Dr. Gio Valiante

Steve Cohen and Jack Nicklaus shared the same performance coach, and for about an hour he's yours too. Valiante's core idea is the “central governor,” the part of the brain wired to keep you comfortable rather than great, and why mastery-focused habits beat ego-driven motivation. Useful listening whether you run a hedge fund or a Sunday four-ball.

🧠 Think: Three Economists Walk Into a Café

The WSJ recently asked three labor economists whether AI will kill jobs, and what struck me is that they read the same evidence and came away with three different answers. No surprise since they are, after all, economists. Anton Korinek sees labor losing its scarcity. David Autor sees work reshaped in several directions at once. Martha Gimbel lands between them. That disagreement may be the most honest answer available right now. 

Some of it isn't in dispute. AI will automate parts of law, finance, accounting, and the rest of what Korinek calls the laptop professions. Obviously, we know that most jobs aren't clear-cut tasks waiting to be optimized. They’re actually messy combinations of judgment, accountability, context, and human interaction. I liked how Gimbel put it when she states that she doesn't care how good the robot gets, she isn't having one raise her toddler. Some work stays human because we decide it should. The transition could still be brutal for experienced workers whose skills lose value faster than they can adapt, which is why the CPO hires I mentioned in the last issue are a good test case on whether experience still holds its value.

But people aren't horses waiting for the engine. Autor's line is that we are the deciders, and I agree with that. AI may reshape the work, but we're the ones who decide what it becomes. Theoretically, at least. 😊

Charts of the Week

The Supply Times Analytics: Millionaire 401(k) accounts are back above 600,000, roughly triple their 2019 level. Before anyone gets carried away, that's still about 2.5% of Fidelity's 25 million accounts, and notice the cliff in 2022 when the market turned. What I'd watch is the timing. This is the wealth cushion sitting underneath the retirement wave we keep talking about, and a healthy balance makes it a lot easier for a seasoned executive to walk away rather than stick around and mentor the bench behind them.

15% of U.S. Employees Use AI Daily

The Supply Times Analytics: Daily AI use at work has risen from 4% to 15% over three years, and more than half of U.S. employees now engage with it in some form. Impressive growth, sure. But hold that 15% next to the section above, where AI-native startups are running with 25% fewer people, and Microsoft is compressing 14 management layers into three. The org charts are shrinking faster than the tools are actually used, which suggests those decisions are being made on promise rather than adoption.

The Supply Times Analytics: This is my favorite chart of the week. The first electric generators appeared around 1880, and productivity growth barely moved for 40 years. The spike didn't arrive until the 1920s, when factories were finally redesigned around electric power instead of just bolting motors onto old steam layouts. That's exactly Koning's point about wiring AI into the workflow rather than handing it to people and hoping.

Quote of the Week

❝

“Attention is the rarest and purest form of generosity”

— Simone Weil

Tweet of the Week

The Final Chuckle

Thanks so much for reading. I’d love to know what you think about this issue and how I can make it more useful to you. If you have suggestions or topics you want to see me address, email me at [email protected]!