Your agent just got a $200,000 budget. It moved $1,000 an hour into a new mix of CTV and retail media while you slept. The first sales readout that would tell you whether any of it worked takes 48 hours to mature. By then, $48,000 has been committed to a decision nobody has assessed, and that first readout will not settle all $48,000 either. It will produce an estimate, with error bars, about a change that is already two days old. AdExchanger columnist Evgeny Popov called this the proof gap in a piece published on October 7, and the gap is the line every brand is about to start crossing without noticing.

Your pacing caps total spend. Your campaign budget caps total spend. Neither one caps how much can ride on a single untested change while the test is still open. Faster measurement narrows the gap but never closes it, because a model updates in milliseconds and a business outcome takes weeks to become real. The gap is structural, and the agentic era just made it the dominant risk in your media plan.

The dollar figure attached to the proof gap

Walk through Popov's example. A brand hands an agent $200,000. The agent mixes CTV with retail media at $1,000 an hour. Forty-eight hours later, $48,000 is committed. That is not the worst case. That is the boring case, because the agent runs every day. In a 30-day month, the same setup commits $720,000 before the first clean readout on the campaign arrives. Most of that money was earned or lost based on a guess the system had no way to test in time.

The math gets worse when the agent changes more than one variable at once. A human media planner can change bids, audiences, and channel mix across a quarter. An agent can do it across an afternoon. The same weak guess now drives spend across several budgets, not one. The gap compounds in proportion to how autonomous the agent is, and the agencies and brands that have spent 2026 granting more authority to their agents have spent that year accepting a larger proof gap on every dollar they touch.

Why faster measurement will not save you

The instinct in 2026 is to ask for measurement that updates faster. Incrementality tests, always-on attribution, real-time dashboards. Each one narrows the window between commit and assess. None of them closes the window, because the business outcome -- a car purchase, a mortgage closing, a brand preference shift -- takes weeks to become real. You can model it in milliseconds. You cannot make it real in milliseconds.

Popov's framing lands here. The standard of proof belongs in the test. The promise to pay belongs in the contract. The cash behind the promise belongs on someone's balance sheet. Three different instruments, three different owners, and treating them as one is the mistake Popov says the industry is about to make on a wide scale. Measurement vendors will pitch faster readouts as the answer. Faster readouts are not the answer. They are a smaller window, with the same gap inside it.

The risk budget is the missing instrument

Popov proposes a second instrument: a risk budget. Alongside the campaign budget, the brand sets the maximum an agent may commit under an unvalidated change before it must show the agreed evidence or obtain fresh approval. An agent gets $12,000 to test a new mix, and nothing more until the readout clears a threshold both sides set in advance. Three rules keep the risk budget honest.

First, all parties agree to the evidence standard before the first bid: what counts as progress, which method measures it, and which parameters need to be met before the agent's scope expands. A pixel recording an event is not the same as a sound causal test with stated uncertainty. The agent's own confidence score is not independent validation.

Second, the clock has to match the goal. A car, a mortgage, and a brand campaign do not prove themselves in 48 hours. A short-term sales gate should not be the default for work designed to pay back over months. When evidence is late, the default is to pause, not to grant fresh funds for an expansion.

Third, the risk budget has to sit outside the agent's objective. A ceiling the agent is rewarded for reaching stops being a ceiling. The agent cannot raise the cap, and it cannot reset the count by moving spend into a new channel or handing it to a second agent.

Who eats the miss when the test fails

The harder question is contractual. If an agent commits $48,000 before anyone sees the readout, and the readout says the change did not work, who pays? This is not a new fight for ad tech. In the early 1990s, the industry argued about sequential liability, the question of whether an agency owed the publisher before the brand paid the agency. That was a credit risk dispute. The agentic version is a proof risk dispute, and the instruments that solve it do not exist in most media buying contracts yet.

Popov predicts that by the end of 2027 at least one media deal between independent firms will settle from escrow, with terms that name the source used to count results and provide a cash remedy for a counting error. The test will be an executed deal with terms anyone can verify. The vendors most likely to enable that structure are not the agencies. They are ad tech insurance products, capital-backed counterparties, and exchange-style pricing for risk transfer. Munich Re is named in the piece tags for a reason. The financial infrastructure for ad tech has been visible since the early 1990s. The industry has not needed it. Agentic buying makes the need obvious.

What to do this quarter

Before you give your next agent more authority, three changes keep the proof gap from swallowing your plan.

First, write down the risk budget. For every agent that touches spend above a threshold you set, document the dollar ceiling on unvalidated commitments, the evidence standard both sides agreed to before launch, and the pause rule when evidence is late. If the agreement lives in a slide deck instead of a contract, the agent will exceed it.

Second, separate the campaign budget from the risk budget in your reporting. The campaign budget tells finance how much you plan to spend. The risk budget tells finance how much you are willing to lose before you have evidence the spend worked. They are different numbers. Treat them as such.

Third, ask your agency or platform who eats the miss. If the answer is "the brand," you are paying for proof risk with no pricing for it. If the answer is "nobody," the agent is running on a default that has not been priced. If the answer involves a contractual counterparty standing behind an evidence standard, you have an instrument most of your competitors do not, and you should be using it.

The proof gap is the line in your media plan that nobody has named yet. Naming it is the first move. Capping it is the second. Putting a counterparty behind it is the third. The brands and agencies that move through those three steps in the next two quarters will write the contracts that the rest of the industry copies by the end of 2027.

Sources