On October 7, Microsoft will hold a Windows and Surface event, its first dedicated hardware showcase in a while. The headlines will be about screens, hinges, and chip specs. The actual story is quieter: Windows stopped being the product years ago. It is the doorway. Every new Surface exists to keep you inside a subscription stack, Copilot included, where the operating system is free-ish and the recurring revenue lives one layer up. Microsoft does not need you to love the hardware. It needs you to stay logged in.
That same logic, infrastructure as the doorway to a recurring toll, showed up twice more this week in very different rooms.
The Cost of Being Simple
A Kitces research note on advisor marketing ROI is making the rounds, and the shape of it is familiar to anyone who has watched a fad chase a fundamental. The advisors winning are not the ones buying the flashiest lead-generation software. They are the ones doing the boring, compounding work: referrals, repeat clients, a reputation built one conversation at a time. The losers spent on paid acquisition channels that produce leads but not trust, and trust is the actual product a financial advisor sells.
This is not a new finding, and that is exactly the point. There is a long and well documented tendency to distrust advice that sounds too simple. A famous story about a consultant who charged a struggling executive ten thousand dollars just to say “write down your six priorities every night” captures it well: the advice worked, and the resistance to it was never intellectual. It was emotional. Simple, unglamorous work does not feel like it deserves the outcome it produces, so people keep buying the complicated alternative instead. Marketing ROI research keeps rediscovering that the plain approach wins, and keeps getting ignored anyway, because it is too plain to justify the budget.
Sales has the same problem now, at a smaller price point. AI sales development reps, the software that qualifies leads and drafts outreach that used to belong to an entry-level hire, are pricing in the low thousands of dollars a year. That is not a rounding error. It is a statement about what a full-time human in that role is now worth relative to a subscription: a fraction of a salary, available instantly, never quitting. The tools are genuinely useful. Automating the repetitive part of sales prospecting is a real improvement, not a gimmick. But the price also tells you how little margin is left in the tasks that used to be someone’s whole job, and how fast that margin compresses once a model can do the task at all.
Whose Copyright Pays for Whose Data Center
Meanwhile, in Australia, a live policy fight is asking a more uncomfortable version of the same question. Regulators are weighing whether to loosen copyright protections specifically to make the country more attractive for AI data center investment. One prominent critic put the trade plainly: it would throw creative workers under the bus. Strip away the politics and the mechanism is simple. A government wants the jobs, tax revenue, and prestige that come with hosting AI infrastructure. The fastest lever available is not tax policy or land use, it is the raw material the models are trained on, and that raw material is other people’s work. Musicians, writers, and illustrators would absorb a policy cost so that a data center could get built with fewer legal headaches.
That is the pattern connecting all three signals. A student who does the reading and shows up with six honest drafts is worth less to a grading system than one who learns to guess the rubric, because the system rewards the signal, not the substance. A sales rep who spent five years learning to read a buyer’s hesitation is worth less than a subscription because the subscription is cheaper to measure. A songwriter’s catalog is worth less to a policymaker chasing data center jobs than the jobs themselves, because the catalog’s value only shows up over decades and the jobs show up in next year’s budget.
None of this is a call to romanticize the old way of doing things. The AI SDR tool is a real productivity gain for a small business that could never afford a full sales team. Microsoft’s ecosystem lock-in has, at least, produced software that mostly works. Cheap infrastructure is how ambitious things get built at all. The uncomfortable part is not that these tools exist. It is that every one of them runs on systematically undercounting the value of work that takes longer to pay off than a quarterly earnings call allows.
We built an entire educational system on the premise that the thing worth measuring is the thing worth rewarding, and grades became the proxy because grades were easy to produce at scale. AI infrastructure is running the same playbook one level up: whatever is easiest to price gets priced, and whatever compounds quietly over years, a songwriter’s back catalog, an advisor’s decade of referrals, a salesperson’s read on a room, gets treated as a rounding error until someone tries to build a business model on top of it and realizes it was carrying more weight than anyone billed for. The boring list still works. It has just gotten harder to convince anyone to pay for the version that takes ten years instead of ten minutes.
Leave a Reply