Running an Independent Tokenary Policy
How to operate a rules-based token policy in practice, so it holds up when the budget meeting gets tense (Series 3/3)
The previous post recommended a destination: skip the easy peg, avoid both loose and tight discipline, and run an independent, rules-based tokenary policy in which token supply grows in step with verifiable business value. That is the right place to arrive. This post is about the harder part - operating it without letting it quietly collapse back into discretion.
It will try to collapse. Every independent policy eventually meets the same moment: a budget is tight, someone influential wants more tokens, and the rule written in a calmer time suddenly feels negotiable. That moment is where a tokenary policy is actually decided.
The Lesson From Uncle Milt Is Pre-Commitment, Not a Formula
Milton Friedman is often remembered for a specific prescription - let the money supply grow at a fixed annual rate - and that particular formula proved too rigid to survive a changing economy. The durable lesson sits underneath it. Policies made by reacting to each short-run moment tend to fail because the effects of any decision arrive with long and variable delays. Cut spending today, and you won't see the damage - stalled projects, frustrated teams - for weeks, by which point you're reacting to a problem you already caused. By the time you see the result, conditions have moved, and your well-meant correction lands as a new disturbance. A series of locally sensible, short-run decisions therefore compounds into instability. The answer is to commit to a rule in advance and hold to it, so that today’s disciplined judgment constrains tomorrow’s more tempted one. A rule decided ahead of time is a defense against your own future weakness - and against the steady accumulation of reasonable-sounding exceptions.
A Single Metric Will Be Gamed
The trouble is that any single number you pick as the target will eventually get gamed. This is Goodhart's Law: once a measure becomes the target, people optimize the measure rather than what it was meant to capture. Set “tokens per resolved ticket” as the goal, and tickets get reclassified to look cheaper. Reward “accepted code changes,” and large changes get split into many trivial ones. The behavior is rarely cynical; teams are simply responding to the incentive they were given. But the result is the same: any single number held up as the target will eventually be optimized into meaninglessness.
Govern With a Basket, Not a Number
The way out borrows from how inflation is measured. No central bank tracks the price of a single good. Instead, it tracks a basket. No one component dominates, so moving the overall figure requires moving many things at once - which is hard enough that the basket tends to track reality despite everyone’s incentives.
The same logic applies to tokens. Rather than govern with a single metric, govern with a small basket of measures chosen so that they pull against one another and evolve as the market evolves. Pick metrics that work for your business. For example, consider the following illustrative approach:
Cost per resolved ticket, set against the rate at which tickets are reopened. A team can close tickets cheaply and badly, but the second measure catches it.
Time-to-deliver code changes, set against post-merge defect or revert rates. Quickness can be inflated with trivial commits, but doing so does nothing for downstream quality and often harms it.
Total token consumption, effective cost per outcome relative to the lowest price for which the same work could be done today. Heavy usage can look productive until benchmarked against what the same outcome would cost on today’s cheapest provider.
The basket is dynamic by design: when a new token provider or open-source optimization drops token costs 30%, swap in a new baseline. Gaming one measure still distorts another, but the whole construct tracks reality instead of yesterday’s pricing sheet.
Let the Budget Move on a Rule
The basket tells the organization whether the value is real. A second rule governs how fast supply may grow. Here, the caution about fixed formulas matters most: per-token costs and model capabilities shift on a scale of months, so a fixed annual budget is very likely to be wrong by the third quarter. The abundance dynamics from the first post mean cheap tokens are consumed faster than anyone expects.
The answer is a rule that adjusts itself rather than a fixed number. Tie the budget to results: token spend may grow in proportion to demonstrated value per dollar - the same value signal the basket is already tracking. When AI genuinely becomes more productive, the budget expands automatically; when it does not, it does not—and no one has to re-litigate the number in a tense meeting. This is pre-commitment doing its real work. The formula is set in advance, against evidence the organization does not control, so that convenience cannot move the goalposts later.
The Three Layers, Put Together
An independent tokenary policy that survives contact with pressure has three layers - one for each problem above: a baseline of reality, a measure of value, and a rule for growth.
A stable anchor - live trends in the cheapest available price for the same work - that the organization rarely touches.
A tensioned basket of measures that resists gaming by design rather than by supervision.
A rule that adjusts itself, letting spending track real productivity without constant renegotiation.
Together, they are the internal central bank the previous post described, made operational.
Key Takeaway
Tokens are an organization’s working currency. Govern them with a basket that resists distortion and a rule set in calmer moments, and the policy will hold when the budget meeting turns difficult. The organizations that capture durable advantage from AI will not be the ones that spent the most tokens or rationed them the hardest. They will be the ones whose rules survived contact with pressure.
Author’s Note
CodifyIQ advises companies on AI adoption - where the token bill, the architecture, and the org chart meet. This series is the thinking behind that work. Subscribe for new pieces as they publish.
This series — Tokenary Policy:
Post 1 — Tokens as the Monetary Supply of the AI Revolution: why cheap tokens lead to bigger bills.
Post 3 — Running an Independent Tokenary Policy (you are here)