Should your online store hire a support agent or use AI?

Black Friday is nine weeks away. Messages are already creeping up, the owner is answering chat at midnight, and the question that has been coming up all year is now urgent: do we hire someone before the peak, or do we let AI handle it?

The honest answer is that it's rarely one or the other. But "a bit of both" is not a plan either. This guide gives you a way to decide based on your own numbers, the factors that matter beyond cost, and what the right first hire looks like when you do make one.

Signs your support needs more capacity

You probably need to change something if:

• The owner or a key person spends hours a day in the inbox instead of on buying, marketing or product.
• Response times grow every week, and evenings and weekends pile up for Monday.
• The same questions take most of the day, while complicated cases wait.
• You're missing sales, because nobody answers product questions when shoppers are online.
• Peaks break you. Every promotion or holiday turns into a backlog that lasts for weeks.

Each of these can be solved with people, with AI, or with both. The right mix depends on what kind of work is piling up.

What AI does well, and what people do well

Work	AI agent	Person
Shipping, returns, sizing and policy questions	Fast, consistent, 24/7	A waste of their time
Order status and tracking	Very good with store integration	Fine, but slow at volume
Product questions before purchase	Good with complete product data	Better for advice and upselling
Complaints and damaged goods	Collects details, then hands over	Essential
Refunds and exceptions outside policy	Shouldn't decide	Essential
VIP and wholesale customers	First reply only	Essential
Spotting patterns ("ten people report the same bug")	Limited	Essential

The pattern is clear: AI takes volume, people take judgement. The question is how much of each your store has.

Step 1: Measure your volume and your mix

Export the last two or three months of conversations. Count:

1. Total conversations per month, and in your busiest month last year.
2. The share that are repeat questions answerable from content or order data (shipping, order status, returns, sizing).
3. The share that need judgement (complaints, exceptions, damaged items, payment problems).
4. When they arrive: how many come outside your current support hours.

These four numbers drive everything below.

Step 2: Work out your cost per conversation

The numbers in this section are illustrative assumptions to show the method. Salaries, productivity and AI prices vary widely by country, store and vendor. Replace every figure with your own.

People. Take the full monthly cost of one support person, meaning salary plus taxes, benefits, tools and management time. Divide it by the number of conversations one person can handle in a month at your quality level.

• Assume a full monthly cost of $3,500 and 1,000 conversations per month.
• Cost per human conversation: $3,500 ÷ 1,000 = $3.50.

AI. Take your vendor's price per AI conversation or per resolution, and estimate the share of conversations the AI resolves without a person.

• Assume $1.00 per AI conversation, and that the AI resolves 50% of them.
• Every conversation goes to the AI first; the unresolved half goes to people.

Step 3: Compare the scenarios

Using the assumptions above, here is the monthly cost at three store sizes:

Conversations a month	People only	AI first, people for the rest	Difference
500	$1,750 (half a person)	$500 AI + $875 people = $1,375	about 20% less
2,000	$7,000 (two people)	$2,000 AI + $3,500 people = $5,500	about 20% less
10,000	$35,000 (ten people)	$10,000 AI + $17,500 people = $27,500	about 20% less

Now change one assumption. If the AI resolves only 30% instead of 50%, the 10,000-conversation store pays $10,000 for AI plus $24,500 for people, a total of $34,500, almost the same as people only. At 70%, it pays $10,000 plus $10,500, a total of $20,500.

The AI's resolution rate decides the business case. Improve your content before you judge the AI, and before you size your team.

Two more things the table doesn't show. Per-resolution pricing changes the maths, because you don't pay for conversations the AI hands over. And at small volumes, the real choice is often not "half a person" but "the owner's evenings", which have a cost too.

Step 4: Look beyond cost

Some of the most important differences don't show up in a cost-per-conversation table:

• Coverage hours. Answering around the clock with people takes roughly four to five full-time people for a single seat, since a week has 168 hours. An AI agent covers nights and weekends for the cost of the conversations.
• Peak flexibility. AI scales up on Black Friday and down in February. People need hiring, training and planning months ahead.
• Speed. AI answers in seconds. People answer in minutes on chat and hours on email, and both slow down in peaks.
• Customer experience. People build loyalty, calm angry customers and sell. A great agent is worth more than their cost per conversation suggests.
• Languages. AI agents can usually answer in the customer's language. Hiring for every market rarely makes sense for a small store.

Step 5: Start with AI on the repeat work

Whatever you decide about hiring, automate the repeat questions first:

1. Fix your help content for shipping, returns, order changes and sizing.
2. Switch on an AI agent on website chat, then social channels.
3. Add an order status flow connected to your store.
4. Read transcripts for two weeks and fix every weak answer.

Then measure what's left. That's the work your next hire will do.

Step 6: Hire for the work AI can't do

The support person a hybrid store needs is different from the one you'd have hired five years ago:

• A problem-solver, comfortable deciding on refunds and exceptions within clear limits.
• Good with upset customers, in writing.
• Commercially minded, able to turn a pre-sale question into an order.
• Curious, spotting patterns in conversations and fixing their causes, including improving the AI's content.

Write the job description around these, not around "answering emails".

When to hire anyway

Even with good automation, hire when:

• the questions AI hands over take more than the time you have, every week, not just in peaks;
• you have important customers (wholesale, VIP, high order values) who expect a named person;
• complaints and returns are growing with your order volume;
• you're opening a channel, like phone, that needs a person.

Common mistakes

• Hiring to answer repeat questions. You pay a person to do what AI does better.
• Judging AI on week one. Resolution rates improve as you fix content.
• Cutting people before AI proves itself. Run both, measure, then decide.
• Forgetting peak months. Model both your cost and your capacity for November and December.

Which setup fits your store

Monthly conversations	Typical setup
Under 300	Owner plus an AI agent and a few automation flows
300 to 2,000	AI agent first, one person part-time or full-time for the rest
2,000 to 10,000	AI agent (managed for you, if you'd rather not run it), a small team with clear roles, and a shared inbox or help desk
Over 10,000	AI agent, a support team with leads and workforce planning, and a full help desk

For an established store in that 2,000 to 10,000 range, price Tidio's Premium plan against your next hire. From 3,000 AI conversations a month, Tidio's own team sets up and keeps improving Lyro and its knowledge, and the plan guarantees a 50% resolution rate with pay-per-resolution billing. It takes the AI workload off your team instead of adding to it.

If you're deciding now, with nine weeks to Black Friday, put an AI agent on the repeat questions first, and if you hire, do it in October so the new person is trained before the peak. Keep your people on complaints and payment issues. Our Black Friday customer service checklist covers the rest, and our guide to AI chatbots for online stores explains what to expect from the AI. For the numbers to track afterwards, see the customer service metrics that matter for online stores.