Black Friday 2026 Customer Support Plan

Last year’s Thanksgiving weekend was the busiest on record. The National Retail Federation counted 202.9 million people shopping between Thanksgiving Day and Cyber Monday, 134.9 million of them online (NRF, 2 December 2025). For this year, Adobe, which measures US online sales directly, forecasts $47.5 billion of online spending over Cyber Week (its five days from Thanksgiving to Cyber Monday), with $12.9 billion on Black Friday and a record $15.1 billion on Cyber Monday (Adobe, 28 September 2026). The NRF’s own 2026 holiday forecast had not been published when we checked on 14 October 2026, so its figures here are 2025 results.

Every one of those orders can come with a question. Will it arrive by the 18th? Does the code work on sale items? Can I return a gift? The useful thing about peak questions is that they are predictable. You can write the answers now, while there is time, and spend the weekend itself on the conversations that need a person.

The countdown below runs from mid October to the returns in January. The dates and figures are American; the plan works the same for a shop anywhere.

The dates that matter in 2026

The USPS dates, from its 2026 holiday shipping page, are 17 December for Ground Advantage and First-Class Mail, 18 December for Priority Mail and 19 December for Priority Mail Express. Your own cut-off comes earlier by however long you take to pack an order in December, and that gap is what customers will ask about.

The questions that dominate the peak

Before the orders, people ask about the deal. After the orders, they ask about the orders. Salesforce’s Cyber Week report shows the second wave clearly: in the shopping data it tracks, customer service conversations handled by AI agents grew 55% week over week, and actions such as updating a delivery address or starting a return grew 70% (Salesforce, 5 December 2025). Here is the usual mix, where each answer should come from, and who should give it.

Six of the eight rows can be answered from something you write once. The last two need someone to look at an order or a payment, and those are the conversations your team should be free for on the day.

Six weeks, one job a week

Today is 14 October 2026. That leaves six weeks before Thanksgiving: enough, if every week has one job and it is finished before the next one starts.

Weeks 1 and 2: find the questions, write the rules

Start with last year. Pull every conversation from November and December: chat transcripts, the support inbox, messages on WhatsApp or social. Many chat tools and help desks export them as a CSV file. Sort them by what the customer wanted, not by how they worded it, and count. The 20 that come up most are your test set for week 4.

Then write the rules down in one place, before marketing has finished the campaign. For each promotion:

• the code, and the exact start and end time, with the time zone;
• what it excludes: sale items, gift cards, bundles, particular brands;
• any minimum spend, and whether it is counted before or after the discount;
• whether it stacks with other codes, with free shipping, or with loyalty points;
• whether you adjust the price for someone who ordered the week before the sale.

Add your shipping cut-offs, your holiday returns window (for example “bought from 1 November, returnable until 31 January”) and your gift card rules. Then publish them as a shipping page, a returns page and a short page of sale terms. A page a customer can read is a page your chat can read.

Week 3: load the answers before the peak

An AI agent answers from what you give it. For the peak, give it three kinds of material.

1. Exact answers for the promotion rules

Codes are where a paraphrase goes wrong: a date off by one, a minimum spend counted before the discount instead of after. Write each rule as a fixed question and answer in the words you want used, so the agent gives that answer rather than composing one. One per code, plus one for stacking, is usually enough.

2. Your shipping and returns pages

The pages you published in week 2 carry the detail: carriers, regions, the returns address, how exchanges work. Add them to the agent, and read them in again every time you change one: until it reads the new version, the chat answers from the old one.

3. A note of cut-off dates

Dates are easier to keep right in a table than in prose. Write them as a CSV file or a short written note, one line per service and region, and replace it when a date changes.

Where is my order? The agent can answer the general part from your published delivery and returns pages: how long dispatch takes during the sale, which carrier you use, and where the tracking link arrives. A specific order is different, because it lives in your order system. With EzyConn, the agent looks one up through an AI Action connected to your own system, from the Starter plan, and only within the rule you set for it.

Stock. Product pages tell the agent what you sell. For questions about what is in stock today, a WooCommerce shop can connect its catalogue (products, prices and stock), as the WordPress and WooCommerce guide explains. A Shopify shop can add its product export as a CSV file and add it again after a big restock; the Shopify install guide covers the rest of the setup.

Week 4: test with last year’s top 20

Testing takes an afternoon and finds the mistakes before your customers do. Use the 20 questions from week 1 and the rules sheet from week 2.

1. 1

   Ask each question as a customer would. On your live site, in a private browser window, typos and all. Do half of them on a phone: Adobe expects a record 63% of Thanksgiving's online spending to come through mobile devices this year.

2. 2

   Read every answer against the rules sheet. Right code, right dates, right time zone, right minimum spend.

3. 3

   Ask the follow-up. Peak conversations are rarely one question. "And does it work on sale items?" is where vague answers show.

4. 4

   Ask something you have not written down. The agent should say it does not know and offer a person, not guess.

5. 5

   Ask for a person. Walk one hand-over all the way through: who is told, where, and what they see when they open it.

6. 6

   Fix it and ask again the same day. Every miss becomes a written answer or a page edit, then a re-test.

Our pre-launch testing checklist has a longer version of this list if you have the time for it.

Week 5: decide what goes to a person

Some conversations should reach a person every time, however busy the weekend is. Write the list down:

• Order problems: damaged or wrong items, parcels marked delivered that never arrived.
• Payment disputes: charged twice, a refund that has not arrived, a declined card that took the money anyway.
• Changes after dispatch: a new address once the parcel is with the carrier.
• Anyone upset, and anyone placing a large or trade order.

Then set your chat up so those conversations reach a person. Two settings do most of it: a list of phrases that mean the customer wants one (“charged twice”, “arrived damaged”, “wrong item”), and exact answers that explain a person will look at it and offer to bring one in. The hand-over should carry the whole conversation, so nobody asks the customer to start again. Our guide to chatbot hand-overs covers what a good one looks like.

Staffing and alerts

Write a rota for Thursday 26 to Monday 30 November. Name one person per shift who takes the hand-overs, and make sure someone who can issue a refund is reachable for payment disputes. Send the alerts where the team already is: if you live in Slack, have each hand-over post there with the conversation attached. If two people are on, share the chats out evenly rather than letting them pile on one.

Last, check your chat plan’s monthly allowance against last November’s volume. Tools meter a busy month differently, per seat, per resolution or per conversation; our guide to what an AI chatbot costs explains each model.

After hours: the evening is the peak

Shoppers buy in the evening, after work. On Cyber Monday 2025, Adobe found that between 8 pm and 10 pm consumers spent $16 million every minute (Adobe, 2 December 2025). For most small teams, that is after the last person has logged off. Here is the same question at 21:10 on Cyber Monday, with and without the rules loaded.

Both lanes are honest; only the second one sells anything. Code rules, cut-offs and returns can all be answered at 21:10. For the rest, three habits help:

• Ask for an email when the chat starts, so a question left overnight still has a way back to the customer.
• Tell customers when the team is back. Put your support hours on the contact page and in an exact answer, so “when will someone reply?” gets a real time.
• Make the overnight hand-overs the first job each morning, before new chats.

Week 6: freeze the rules

By 18 November, get the final codes from whoever runs the campaign, and write the last exact answers. Read every changed page into the agent again, then run the top 20 one final time. After that, treat the rules as frozen: a new code or a changed end time goes live only with its written answer the same hour. A late change that never reached the answers is the easiest peak mistake to make.

On the day: a 20 minute morning check

1. 1

   Check what it could not answer overnight. Read the hand-overs and the reasons in the reports, and write an answer for any question that came up more than once.

2. 2

   Reply to the overnight hand-overs. Oldest first. They were promised a reply.

3. 3

   Look at yesterday's first reply times. If people are slow on hand-overs, move someone onto the chat for the evening.

4. 4

   Look for a new cluster. A code that will not apply, a carrier delay, a product that sold out. Write one answer and put it on the site too.

What to measure, on the day and the week after

The gaps are the line to act on. Each question you answer before 1 December is one fewer in the run-up to Christmas.

After the weekend: cut-offs, then returns

December is a month of changing answers. The day after each cut-off passes, update the exact answer: once standard delivery can no longer arrive before 25 December, the agent should say so, name the faster service that still can, and mention the e-gift card. Put a reminder in the calendar for each date in your cut-off note.

Then the returns arrive. In the NRF and Happy Returns survey of large US retailers, merchants expected 17% of holiday sales to come back, and 19.3% of online sales across the year (NRF, 15 October 2025). The questions are predictable: how to return a gift without the order email, how exchanges work, and when the refund lands. Write those answers in October with everything else, so 26 December is just another day on the rota.

Where EzyConn fits

EzyConn’s AI agent answers customers on your website chat, WhatsApp, email, Telegram and an inbound phone line, from your own content, in multiple languages. The plan above maps onto it directly:

• Your content: give it your store’s address and keep the pages you want; “read it all again” re-reads them in one click after an edit. Add the cut-off note or a Shopify product export as a CSV file, and write each code’s rules as an exact question and answer, which is used first.
• When it does not know, it says so and offers a person. When a teammate answers instead, that reply waits as a draft answer for you to approve, and the Unanswered screen shows the topics it answers least.
• The hand-over: list the phrases that mean “get me a person”. When the AI offers a person and the customer accepts, the AI steps back and a teammate is assigned and notified in the inbox and in Slack (or Microsoft Teams (beta)) with the whole conversation. Chats can be shared out round-robin, to the least busy or the last active teammate, or by hand, and “Hand back to AI” returns the conversation afterwards.
• Orders: AI Actions, from Starter, let the agent look an order up in your own system within the rule you set.

Pricing is per workspace, with no charge per resolution, so a busy weekend costs the same as a quiet one within the plan’s allowance. The Free plan includes the AI agent, needs no card and covers 500 messages and 1,000 AI credits a month; the pricing page lists every plan’s allowance. How EzyConn works walks through the hand-over step by step, and EzyConn for e-commerce shows the setup for a shop. You can start free and ask it last year’s top 20 this week.

Black Friday customer support: common questions

Keep reading

• Handling returns and refund questions with an AI chatbot
• After-hours support with an AI chatbot
• Ecommerce live chat: the four jobs it has to do
• A free AI chatbot for a Shopify store
• AI customer support with a person when it matters