Does Digital Wallet Stop Agentforce When Credits Run Out, or Do We Need Our Own Kill Switch?

An alert is not a control. Digital Wallet tells you the meter is running. It does not turn the meter off.

That distinction decides whether your first hard stop is a threshold you chose or an invoice finance did not expect. Salesforce is direct about the gap, which makes this a solvable operations problem rather than a vendor argument.

This page is the control plane. It is not the Flex Credits forecast-and-close model, it is not a partner product review, and it does not invent a Salesforce list price. Every product claim below carries the URL that published it. Your order form beats any blog math, including ours.

The buyer question, stated cleanly

Director of RevOps, revenue finance partner, Salesforce admin, VP Sales Ops at 100+ headcount:

If Agentforce burns through the credit balance overnight, does anything stop it, or do we find out at close?

Useful answer: nothing in the platform stops it. Enforcement is a rule you write, an owner you name, and a switch a human throws. Build that before the cohort scales.

What Digital Wallet actually does

Salesforce positions Digital Wallet as near real-time consumption insight across Flex Credits and Agentforce usage, with alerts so teams can see burn as it happens rather than at invoice time. Source: Salesforce.

That is a visibility layer, and it is a good one. Most consumption vendors do not ship anything comparable. The mistake is reading visibility as governance. A dashboard that updates every few minutes still requires a person to act on it.

What Salesforce itself says about the gap

Salesforce's 18 August 2026 post on the Agentforce usage data model, by Chetan Dandekar, states the critical line plainly: Digital Wallet does not block execution when credits run out, and overages are billed at contracted rates. Alerts are early warning, not enforcement. Source: Salesforce.

The same post gives the setup path. Go to Setup, then Flow, and activate the Consumption Threshold Alerts template at 75 percent, 90 percent, and 100 percent of the balance. That is a fifteen minute admin task with no license implication. Source: Salesforce.

One more point from that post matters at close. The token data model is directional for analysis. Digital Wallet is the source of truth for billing. If your usage report and your invoice disagree, the invoice is the record. Source: Salesforce.

Salesforce Ben reaches the same conclusion from the buyer side: Digital Wallet provides tracking and overage alerts but has no kill switch to freeze agents when a threshold is hit, and a 10,000 token per action cap means one task can be counted as multiple actions. Source: Salesforce Ben.

Why alerts alone create a Proof Gap

Three mechanics compound.

1. The invoice is the first hard stop. Nothing between the alert and the bill interrupts execution, so the ceiling is financial and retrospective rather than operational and immediate.
2. Token-split actions break the unit forecast. A verbose prompt template turns one designed task into two or three billable actions, so burn outruns the model without anyone changing the use case.
3. Finance sees it late. The month you overspend is the month you report as a success. The correction lands in the next cycle, against a period the team has already closed.

That is the Proof Gap in its cheapest form. The claim was built on an in-month dashboard. The receipt arrives weeks later and nobody restates the number.

The control plane before production scale

Five items. None of them require a purchase.

1. Alert thresholds live. Activate the Consumption Threshold Alerts Flow template at 75, 90, and 100 percent, and route the notification to a channel a human watches, not a mailbox nobody opens.
2. A weekly Digital Wallet owner. One named person in RevOps, same day each week, logging burn to date, run rate, and projected period end.
3. A written kill rule with a named admin. State the trigger, the action, and the person. Example shape: if projected period-end burn exceeds the cohort ceiling by a set percentage for two consecutive readings, the named admin disables the agents in that cohort until the use case is re-scoped.
4. A cohort ceiling you enforce yourself. The platform will not cap a cohort, so cap it in your model and hold the owner to it. One cohort, one number, one owner.
5. Optional partner tooling. Buying an accelerator can shorten the response loop. It does not replace the rule, the owner, or the ceiling. Do not let a purchase decision delay writing the rule.

Test it once before scale. Trip the 75 percent alert deliberately in a sandbox cohort and time how long it takes the named admin to disable an agent. If that number is measured in days, the rule is decorative.

Who is selling into this gap

Demand for a control plane is visible in the partner market. Simplus published two Agentforce Kill Switch Accelerator posts on 24 and 27 August 2026, arguing there is no native per-action kill switch and offering closed-loop shutoff tooling. Sources: Simplus and Simplus.

Cite that as evidence the question is real, not as the answer. A packaged shutoff is a faster hand on the switch. The rule, the threshold, and the owner are still yours to define, and they cost nothing.

What this is not

This is not an anti-Agentforce page. Publishing the gap and shipping the alert template is more transparency than most consumption vendors provide.

This is not the Flex Credits forecast-and-close model. That page covers burn modeling, arrears invoicing, and the close checklist. Read it at How Do You Forecast Agentforce Flex Credits When Overages Bill in Arrears.

This is not a Simplus product review. We have not tested the accelerator and take no position on it beyond citing it as market evidence.

This is not a price list. No Salesforce list price appears on this page. Confirm your rate card version and contracted rate on the order form before anything reaches a budget slide.

Related: Agentforce Flex Credits forecast and close · Claudeforce two contracts, two meters · AI SDR kill criteria · The Reversal Ledger · Agentforce and Data 360 share one credit pool