An AI agent can help buy Black Friday and Christmas media, but only once it has a written mandate, hard spending limits and a named person who signs off on anything it cannot undo. Peak season is the worst time to hand an untested agent new freedom, because spend concentrates into a few days and mistakes compound before anyone can correct them. For most advertisers, the sound answer is to use agents for monitoring, recommendations and drafting in Q4, and to allow automatic action only for tasks the agent has already done well within limits.

TL;DR: Use an agent for peak-season media only if it has been tested on ordinary weeks first, and fix its mandate, budget ceilings and approval thresholds before November rather than during Cyber Week. Keep any commitment that cannot be reversed, such as a guaranteed deal or a new seller, behind human approval. Judge the agent on the terms it secured, not on whether it finished the buy.

Should you use an AI agent for Black Friday and Christmas media buying?

Yes for some jobs, no for others. The deciding factor is less the agent's capability than how much of the season's money sits in days when there is no time to fix an error.

That concentration is large. US consumers spent $257.8 billion online between 1 November and 31 December 2025, and the five days from Thanksgiving to Cyber Monday brought in $44.2 billion of it, according to Adobe's 2025 holiday season results. For 2026, Adobe forecasts $275.1 billion online, with those same five days expected to drive $47.5 billion, or 17.3% of the total, as reported by Adweek.

The UK pattern is similar. The Advertising Association/WARC Expenditure Report forecast UK ad spend of £12.0bn in Q4 2025, up 7.3% year on year, with online formats taking 83% of it. An agent that drifts from its brief for a few hours in March costs little, but the same drift on Black Friday lands on the most contested inventory of the year.

Why is peak season harder for an AI buying agent than the rest of the year?

Peak season changes the conditions an agent was tested in. Sale windows have stretched: a media buyer told Digiday in November 2024 that brands joining Amazon's Black Friday event that year had to run the deal for 12 days, from 21 November to 2 December, rather than the few days many had planned. Pacing logic built for an average week then has to cope with a different shape of demand, mid-flight.

Offers, creative and available inventory also change daily in Q4. Each change is a moment where an agent either follows a rule it was given or improvises, and improvisation is where unplanned commitments come from.

Peak season also leaves no room to learn. A flaw found in October is a useful test result, while the same flaw found in the first week of December is an incident. The practical response is rehearsal before the season, in conditions that imitate peak without real money at stake.

As a worked example, the Alkimi Gym lets buyers and sellers build their own agents, set the rules they work to and spar against counterparty agents before any real money moves. Whatever tool is used, the principle holds: an agent's first high-pressure week should not be a live one.

What should be in place before an agent touches peak-season budgets?

The checklist below runs in four phases, from late October to January. Each item can be ticked off by a named person, and any item left open is a reason to keep the agent at a lower level of autonomy.

Phase 1: before the end of October

Write the mandate down. State the objective, channels, audiences, exclusions and who the agent acts for, using a structure such as this media buying mandate template.
Set daily and hourly ceilings, not only flight budgets. A pacing error on one day should not be able to spend a week's money.
Define approval thresholds. Decide which commitments need human sign-off by value, deal type and seller, following the logic in how to design approval thresholds for agentic campaigns.
List what the agent may never do. Typical exclusions are signing guaranteed deals, adding new sellers, changing brand safety settings and accepting terms outside the mandate.
Name the people. Record who approves, who can pause the agent and who covers weekends, Thanksgiving and the UK bank holidays.
Fix the minimum terms of a good buy. Agree the floors for viewability, make-goods, frequency and brand safety before the agent negotiates, not after.

Phase 2: before Black Friday week

Run the agent on ordinary weeks first. Compare its plans and outcomes with what the human team would have done.
Rehearse peak conditions. Test price spikes, sold-out inventory and an offer that changes mid-flight.
Test the pause. Stop the agent deliberately and confirm how quickly spend actually stops.
Check the deal record. Every agreed deal needs a written record of terms that both buyer and seller can retrieve.
Agree escalation routes with sellers. Ask key publishers how they handle agent-negotiated deals and whom to contact if terms are disputed.

Phase 3: during peak

Review agent decisions daily. A weekly review is too slow when about a sixth of the season's online spending lands in five days.
Watch pacing against each day's plan. Early overspend and late underspend both show up here first.
Hold autonomy at its October level. Do not promote the agent to more independence mid-season, however well it seems to be doing.
Log every override. Each human correction is evidence for next year's mandate.

Phase 4: after Christmas

Compare promised terms with delivered terms. Check each deal's viewability, make-goods and frequency against what was signed.
Plan for the January sales. The season does not end on 25 December: Publicis Groupe UK's 2026 study puts UK consumers' planned spending at £17.9bn across the Golden Quarter retail window, from Black Friday to the January sales.
Decide the agent's next step on the evidence. Widen its permissions only where the log shows it stayed within limits.

Which peak-season decisions should stay with a human?

Any decision that cannot be reversed should stay with a person during peak. That covers guaranteed or upfront commitments, new sellers, contract terms below the agreed floors, brand safety changes and spend above the approval threshold.

A useful frame is earned autonomy, a five-stage ladder of Observe, Recommend, Draft, Human-approved action and Bounded automatic action (what earned autonomy is). For peak season, a sensible rule is to keep the agent at the stage it reached by October and not one higher. Monitoring, pacing alerts, reporting and draft proposals sit comfortably at the lower stages.

Early adopters describe the same instinct. Sydney Brower, director of digital media and strategy at the agency InterMedia, told AdExchanger in October 2026 that she prefers to check inputs in each DSP dashboard before giving an agent the go-ahead: "You can trust AI to fly a plane, but I still want a pilot in the cockpit to watch it."

How do you know the agent did a good job, not just a fast one?

Finishing a buy and buying well are different tests. In Concourse Bench v1, a benchmark of AI media-buying agents, 55 of 96 registered attempts ended in a completed buy, but 95 of the 110 contracts signed in those buys promised a viewability floor below 70%, according to Concourse.

For peak season, that gap matters more than usual. An agent that closes dozens of deals during Cyber Week may have closed them on weak terms, and no one may look until January. Reviewing the signed terms, not only the cost and the outcome, is the check that catches it.

Where do industry standards fit for Q4 2026?

Standards are worth asking vendors about, even though they will not change this year's checklist much. AAMP, IAB Tech Lab's Agentic Advertising Management Protocols, is an umbrella initiative for bringing agentic AI into advertising, built on the standards and taxonomies the industry already uses. Its core OpenProposal specification, which defines how buying and selling agents exchange proposals, is open for public comment until 10 November 2026 after IAB Tech Lab extended the window on 23 October, as IAB Tech Lab reported.

AdCP, the Ad Context Protocol published by AgenticAdvertising.org, describes itself as an open protocol that lets AI agents discover, plan, buy, sell and measure media across channels. For a Q4 buyer, the useful question is which standards a vendor's agent works with and how its deal records can be checked.

What does this mean for planning Christmas 2026?

The advertisers likely to get value from agents this Christmas are the ones who settle the rules in October and use Q4 to gather evidence rather than to experiment. Agents can take on the watching, reporting and drafting that peak season multiplies, while people keep the commitments that cannot be undone.

Alkimi, which runs the Concourse benchmark cited above, builds its Marketplace on the same principle: humans keep approval rights at the deal level, and each agreed deal is written to a DealSheet that both parties hold and can retrieve independently. Whichever route an advertiser takes, the checklist above is the part worth finishing before Black Friday.

Frequently asked questions

Should I let an AI agent run my Black Friday campaigns on its own?

Not unless it has already run comparable campaigns within limits on ordinary weeks. Peak season concentrates spend into a few days, so irreversible commitments should stay behind human approval.

What can an AI agent safely do during Q4?

Monitoring, pacing alerts, reporting and drafting proposals are low-risk jobs for an agent during peak. Bid and budget changes within hard daily caps are reasonable once the agent has a track record.

When should I start preparing an agent for Christmas?

Start before the end of October, so the mandate, caps and approval thresholds are fixed and tested before Black Friday week. Changing the rules mid-season makes it hard to tell whether a problem came from the agent or from the change.

What is the biggest risk of using an AI agent in peak season?

The biggest risk is an error that runs unnoticed on the most expensive days of the year. A tested pause and a daily review of the agent's decisions limit how far that error can spread.

How do I measure whether the agent did well over Christmas?

Compare the terms each deal promised with what was delivered, alongside cost and outcomes. An agent that completed every buy on weak terms has not done the job.

Should an AI Agent Buy Your Peak-Season Media? A Q4 and Christmas Checklist