AI Foundations Pro+ ~8 min read New · July 2026

AI spend control: the meter is running.

AI spend starts as one harmless subscription and grows into a line item nobody audits — flat fees, per-seat plans, and now usage meters that scale with enthusiasm. The control system is simple; the discipline of running it is the Pro+ part.

01 The inventory (most people can't produce one)

One page, right now: every AI subscription (personal AND business cards — they hide on both), every seat you're paying for others, every usage-billed API or agent, monthly cost of each, and the lane it owns from your stack card. Most people miss at least one forgotten subscription on the first pass — that find alone usually pays for this lesson's tier.

02 The three billing shapes, three disciplines

Flat subscriptions: the hour test, monthly — did it save an hour of your loaded time? Miss twice consecutively → cancel or downgrade. Sentiment is not a metric; the calculator is.
Per-seat plans: the usage audit, quarterly — admin dashboards show who's actually active. Idle seats are pure waste wearing a team-spirit costume; reassign or drop them without ceremony.
Usage meters (APIs, agents, Work-style tasks): the dangerous shape, because cost scales with enthusiasm and nobody feels cents accumulating. Non-negotiables: budget caps and alerts BEFORE the first real run, a rough cost-per-job number for each pipe, and a monthly glance at the meter. Every usage-billing horror story is a missing cap.
The honest denominators

Two numbers make AI spend rational: cost per hour saved (total spend ÷ honest hours saved — under your loaded hourly rate means the stack is profitable) and spend as % of revenue for owners (context: it should be a rounding error next to what the saved hours earn — if it isn't, the stack has passengers). Run both quarterly; decisions become obvious.

03 The quarterly true-up (30 minutes)

Calendar it with your other quarterly reviews (memory audits, agent-fleet review, red-team — the Pro+ maintenance suite travels together): update the inventory, run both denominators, kill or downgrade the failures, and check for consolidation — this market bundles fast, and the feature you pay one vendor for keeps appearing free inside another (this quarter's version of that story is on our what's-new page, perpetually). Spend that's reviewed stays a tool; spend that isn't becomes a habit with a credit card.

Try it now

Build the inventory tonight — both cards, all seats, every meter. Circle anything that would fail the hour test this month. That circle is your true-up agenda.

Run your numbers →

This week's challenge

Full control system this week: inventory built, caps and alerts on every meter, both denominators computed, quarterly true-up on the calendar. Then the satisfying part — cancel the first thing that fails the hour test, and put the savings toward the tier where you're reading this.

Up next in AI Foundations

How AI actually works

The loop closes — the mental model under everything. Read the lesson →