Start in October, or early November at the latest. Almost everything customers ask can be predicted: stock, delivery dates, offer rules, order tracking. Set up well, an AI agent sells more: it answers pre-sales questions, simple or very specific, within seconds. Your team keeps the after-sales cases that need judgment and action. Then plan for what comes after: the following days, and January returns.

Black Friday: what the AI agent does and what the team keeps, from October to January October get ready Black Friday week Nov 23 to 30 Days after early December January returns The AI agent takes Offer rules Event mode Dress rehearsal Pre-purchase Stock, delivery Promo codes Where is my order Package tracking Delays Return policy Exchanges Return tracking Your team keeps Writes the answers Sets the limits Watch plan Upset customers Exceptions Goodwill gestures Stuck packages Carriers Second contacts Off-policy returns Refunds Special cases
From October to January: what the AI agent handles and what your team keeps, at each step.

When is Black Friday 2026, and when should you get ready?

Black Friday 2026 falls on Friday, November 27, the day after Thanksgiving in the US, and Cyber Monday on Monday, November 30. Many brands open their deals earlier, at the start of the week or even the week before: in 2025, Amazon started its "Black Friday Week" on November 20 (Amazon announcement, checked in October 2026).

For customer service, the time to prepare is October and early November, while your team still has time to write, test and fix things. During Black Friday week, nobody will.

If you're putting an AI agent live for the first time, do it several weeks before the peak. Its first weeks are when you spot what it doesn't know yet, and that work is easier when things are calm.

What do customers ask during Black Friday?

The same things as the rest of the year, in much larger numbers, in two clear groups.

Before buying, the customer is hesitating and looking for a reason to order now:

  • is it in stock, in my size, in my color?
  • will it arrive before the holidays if I order today?
  • can I stack this code with the discount shown on the page?
  • can I pay in installments?
  • what's the difference between these two models?

After buying, the customer is waiting for the package:

  • where is my order? Support teams call this WISMO, for "where is my order".
  • can I change my address or cancel?
  • how do I return an item?

In practice, most of these questions have an answer you know in advance, or one that can be read from the order data. So they're easy to hand to an AI agent, as long as it can see that data. An agent that sees neither stock nor orders can only send people to your team, and the wave lands back on them.

Does an AI agent cost you sales on Black Friday?

No. Set up well, it wins you sales. It clears up doubts while the customer is deciding, within seconds, evenings and weekends included, when your team is offline.

This is the part that matters. The usual fear is that an AI agent handles customers worse than a person would, and that sales slip away. The fear has a basis: an agent that can't see stock, or gives last month's delivery times, can cost you a sale. But with a well-connected agent, I see the opposite at the brands we work with: it sells more. I wanted to be sure of it, so we track the orders placed after a conversation with the agent, and the ones where it really helped.

Before buying, customers rarely hesitate over a simple question. They want to know whether the size will be right, whether two discounts stack, whether the parcel will arrive before the holidays. An agent connected to the catalog, stock and delivery times answers those specific questions as well as the simple ones, within seconds, at any hour.

The customer still has to ask. A short message on the site, inviting them to ask whatever is holding them back, brings in customers who wouldn't have thought of writing. At one of the brands we work with, an invitation like this has been running since September. Customers who buy after talking to the agent, and who write from the website, are usually chatting from a product page: dimensions, delivery, material, which of two models to pick, paying in installments. Most of them order within a day of the conversation.

"Set up well" means three things: the agent reads the live catalog, stock and delivery times instead of reciting a product page it learned weeks ago, it knows the exact rules of your offers, and it hands over to a person, with the context, when a question goes beyond what it can do. If one of them is missing, the agent gets in the customer's way when it should be helping them decide. You can see how we build this kind of agent on our AI agent page.

Six things to prepare before the peak

Prepare the agent in six steps, during October and early November.

  1. Give it the offer rules before launch. Dates, products included, how discounts combine, code conditions, guaranteed delivery dates for the holidays. Add them to its knowledge from the first day of the offer, so it answers "how do I use my code?" and "can I combine it?" correctly.

  2. Prepare the invitation to ask. A short message on product pages, matched to the offer (for example "A question about delivery before the holidays?"), that shows without getting in the way of the purchase and that you can switch off at any time.

  3. Plan an "event mode" that switches itself off. Peak-time instructions are temporary: a tone for first-time visitors, the details of the offer, current delivery times. Plan them in two phases, during the peak and in the weeks after, with an end date that switches them off. Nobody has to remember to remove them, and the agent can't keep talking about an offer that's over.

  4. Run a dress rehearsal. Before the day, send the agent a large number of simulated conversations at the same time, and check that it still answers quickly and that handovers reach the right people.

  5. Decide what the agent never does on its own. Discounts, apology codes, refunds, goodwill gestures: these decisions commit the brand. At an online retailer preparing for a traffic spike, an apology code was considered, but it stayed switched off until the brand's team approved it, and the agent couldn't hand one out on its own.

  6. Write down who watches what during the peak. Who reads the handed-over conversations, who checks that the agent is still answering, who can switch a feature off if a tool slows down. One page is enough, as long as it's written beforehand.

Pre-sales, after-sales: who does what during the peak

On pre-sales, the agent handles about nine conversations in ten on its own, simple questions and very specific ones alike. That's what we measured in October 2026 at the brands we work with, over a month of conversations sorted by an AI. I don't buy the split you often hear, AI for the simple cases and the team for the complex ones: give the agent the right data and it answers a detailed question about a product or an offer very well.

The real line runs between pre-sales and after-sales. After the purchase, your team often has to make a call, then act in its own tools. The agent hands over for:

  • failed or stuck deliveries, which need a decision with the carrier,
  • damaged or wrong items,
  • exceptions to your rules and goodwill gestures,
  • upset customers, especially those writing for the second time,
  • sensitive requests: disputed payments, complaints, possible fraud.

And it hands over properly: the customer doesn't have to explain everything again, because the agent passes on the request, the order involved and what it has already checked.

Plan for the moments when nobody's available, too. A message that promises a reply with an honest delay and opens a ticket reassures people far more than a bare "our team is unavailable".

Why does a second peak come after Black Friday?

Because customers who ordered over the weekend start waiting for their package on Monday. "Where is my order?" messages climb over the following days, especially when carriers, also swamped, fall behind. Then come the questions about delivery before the holidays, and in January the returns and exchanges.

For that second wave:

  • connect the agent to order and shipment tracking, so it gives the real delivery status instead of a generic link,
  • prepare answers about carrier delays, with what your brand offers,
  • review your return policy before January, and give the agent its exact wording.

If time is short, start with two things: the offer rules and shipment tracking. Those are the questions the agent will get no matter what.

FAQ

Should you launch an AI agent right before Black Friday?

It's better to do it several weeks before. The first weeks are for seeing what the agent doesn't know yet and filling the gaps. During Black Friday week, your team won't have time for that.

Can an AI agent handle pre-sales questions on its own?

Yes, about nine in ten at the brands we work with, as long as it's connected to the live catalog, stock and delivery times. It answers simple questions as well as very specific ones (dimensions, compatibility, stacking discounts, delivery before the holidays) and hands over when the team needs to make a decision.

Can an AI agent give a discount or a promo code?

Technically yes, but we don't recommend it without strict rules. A discount commits the brand and is decided case by case. The agent can explain current offers and their conditions, and pass a request for a goodwill gesture to your team.

What happens if the AI agent can't answer during the peak?

It should hand over to a person with the context of the request, or open a ticket your team will pick up as soon as it can, with an honest delay. A customer who knows when they'll hear back waits much more patiently than one left without news.

Is Black Friday one day or a whole week?

The day itself is Friday, November 27, 2026, but many brands spread their deals over the whole week and through Cyber Monday on November 30. For customer service, the load lasts even longer, with the delivery questions that follow.