← All field guidesBudget policy · Design limits

One site-wide AI limit or separate limits by feature?

Customer support, content generation, and internal automation share one budget but have different business value. Decide whether simplicity outweighs the risk that one low-value feature exhausts capacity for another. This budget-policy design gives the concrete numbers, evidence, failure mode, action order, and completion test needed to make that decision responsibly.

Updated 2026-08-17 · 4 min read
Written for
WordPress technical lead
Article format
Decision comparison — One site-wide AI limit or separate limits by feature?
Take-away
Feature-Level Limit decision matrix

Name what each option protects: One site-wide limit versus limits by feature

Customer support, content generation, and internal automation share one budget but have different business value. A useful comparison for one site-wide limit versus limits by feature begins by stating the legitimate value of each option instead of constructing one obviously bad alternative.

With one $100 ceiling, a nightly content job can spend $90 before a morning support assistant opens, leaving a high-value customer path only $10 of capacity. Apply both options to the same numerical case for one site-wide limit versus limits by feature, including downtime, exposure, customer outcome, and operator effort.

Compare one real scenario: One site-wide limit versus limits by feature

For each feature, record normal and peak cost, value per successful outcome, deferrability, fallback, failure pattern, owner, recovery priority, and the operational cost of maintaining another rule. For one site-wide limit versus limits by feature, define a success signal and a failure signal for every option so the chosen policy can be evaluated after adoption.

Decide whether simplicity outweighs the risk that one low-value feature exhausts capacity for another. The operating boundary is explicit: Use one limit when functions have similar value and recovery priority; split limits only when the stop order, owner, or customer consequence genuinely differs. If one site-wide limit versus limits by feature remains a close call, prefer the option that preserves more evidence and can be reversed with less customer harm.

  • Evidence set — For each feature, record normal and peak cost, value per successful outcome, deferrability, fallback, failure pattern, owner, recovery priority, and the operational cost of maintaining another rule.
  • Decision boundary — Use one limit when functions have similar value and recovery priority; split limits only when the stop order, owner, or customer consequence genuinely differs.
  • Completion check — Can the owner of one site-wide limit versus limits by feature state the condition that would justify switching to the other option?

Price the wrong decision: One site-wide limit versus limits by feature

Creating a separate limit for every hook can make the policy harder to operate than the risk it controls, while one undifferentiated pool lets low-value automation starve customer service. A feature checklist fails for one site-wide limit versus limits by feature when it ignores the consequence of a false stop, an overage, or an operator unable to recover the site.

Inventory functions; group those with the same business consequence; compare shared and separate designs; test exhaustion; verify fallback and recovery order; assign ownership; revisit after one billing cycle. Use the decision order for one site-wide limit versus limits by feature to move from customer job to exposure, evidence, pilot, and review rather than choosing from plan or feature labels.

Choose the reversible test with Feature-Level Limit decision matrix: One site-wide limit versus limits by feature

For one site-wide limit versus limits by feature, Free enforces sitewide and per-source monthly estimated-USD, request-attempt, and total-token limits. Pro 1.5 adds rolling one-minute and one-hour token limits, repetition and retry windows, per-request controls, source-aware burst handling, provider/model caps, and context policies. A daily USD cap or a custom multi-signal short-window rule described here still needs external monitoring or application logic.

Use the matrix to document why a function is grouped, not merely where a setting lives, so architecture changes do not erase the business priority. Keep the Feature-Level Limit decision matrix connected to the provider's final bill because model pricing, discounts, caching, and currency conversion can make an operational estimate differ from the amount ultimately charged.

Record when to reconsider: One site-wide limit versus limits by feature

At the review date, compare the observed success and failure signals for one site-wide limit versus limits by feature with the ones written before implementation. The completion question is: “Can the owner of one site-wide limit versus limits by feature state the condition that would justify switching to the other option?” Record the answer, the remaining uncertainty, the owner, and the next review date rather than treating an executed action as a completed outcome.

The Feature-Level Limit decision matrix should preserve why one site-wide limit versus limits by feature led to this choice, what the rejected option still does well, and what evidence would reopen the decision. For one site-wide limit versus limits by feature, that record creates a natural next step: test the chosen boundary on one supported, reversible WordPress path, confirm the customer fallback, and expand only when the evidence still supports the decision.

Download AI Cost Guardrails-CNXT and turn the compatible parts of “One site-wide AI limit or separate limits by feature?” into live WordPress protection. Free enforces sitewide and per-source monthly USD, request-attempt, and token boundaries; Pro adds supported short-window and failure-pattern controls.

Next field guideWhy a monthly limit needs a daily guardrail →