6 Things an AI Chatbot Handles After Your Team Goes Home

Your business has opening hours. Your customers mostly don't care.

When we launched a chatbot for a REIT-owned centre in the New York metro, with 90+ stores, we expected the conversations to follow the doors. They didn't. Over the first nine days, 28% of conversations arrived outside staffed hours. The earliest came in at 7:18 in the morning. The latest at 11:06 at night. Some arrived on a day the centre isn't open at all.

That is more than a quarter of customer questions landing when nobody was there to answer. What an AI customer service chatbot does with those questions is more useful than most businesses expect.

1. It answers the people who plan ahead

After-hours visitors are rarely in a hurry. They are planning tomorrow's trip, the weekend errand, the gift they need by Friday.

On another shopping centre deployment of ours, some of the most useful conversations were late-night planning sessions, held hours after the last store had closed.

Without an answer, that person either guesses, checks a competitor, or forgets. With one, they arrive already decided.

2. It speaks the languages your team doesn't

The New York centre's chatbot handled English and Spanish without being asked to. In one conversation, it walked a shopper turn by turn between two stores entirely in Spanish.

Our other shopping centre deployment answered accurately in Spanish and in Hungarian.

Very few businesses can staff every language their customers speak, and almost none can do it at 11 at night. A well-built chatbot answers in whatever language the customer opens with. For a lot of your customers, that is the first time your business has spoken to them in their own language.

3. It absorbs the routine questions so your team doesn't have to

Hours. Parking. Where a store is. Whether something is in stock. Holiday closures.

These questions are not glamorous, and they are most of the volume. The New York centre's chatbot handled 92 conversations in its first nine days, about ten a day, on a widget nobody had advertised. The other deployment handled 340 conversations in its first 27 days, and two thirds of them were high-intent.

Every one of those is a question your front desk or phone line did not have to answer. More importantly, every one got an answer in seconds instead of on the next business day.

4. It routes the urgent ones to a person

Not every after-hours message is routine, and a good chatbot knows the difference.

At the New York centre, a shopper reported a security incident after hours. The chatbot routed it to mall security and to the management contact form within 30 seconds, with no staff on site.

This is the part most businesses skip when they set a chatbot up. Before launch, decide which messages should never be answered by a machine. Safety issues, complaints, anything involving money or health. Then decide exactly where each one goes and who gets notified. A chatbot without escalation paths is a liability. A chatbot with them is a night shift.

5. It hears demand you had no other way to hear

Customers tell a chatbot things they would never bother telling a person.

The New York centre's chatbot produced a ranked list of brands shoppers asked for that the centre does not carry. That list included unmet demand in a category the centre already believed it had covered. One tenant that had not opened yet drew seven separate enquiries in nine days, roughly 8% of all conversations. That is a measurable demand curve for a single store before it opened its doors.

We wrote about this side of chatbots in AI for Retail: The Shopper Data Hiding in Your Chatbot. The short version is that service and intelligence come from the same conversations.

6. It needs a manager, like any member of staff

A chatbot is only as accurate as what it has been told. Stores change, hours change, services change, and the knowledge base drifts out of date unless someone owns it.

That is why monitoring is part of every chatbot we manage, not an add-on. Reading what customers actually asked shows you where an answer was incomplete or out of date, before a customer notices. It also shows you which questions keep coming up, which is usually the first sign your website is missing a page.

Set up and walk away, and accuracy slips. Set up and review monthly, and it keeps improving.

Before you choose a chatbot

Ask three questions. Can it route urgent messages to a real person, with a notification? Will someone review the conversations every month? And can you see what customers asked, organized, rather than scrolling a raw log?

If a vendor can't answer all three clearly, keep looking. We covered more of these in Top 6 Questions to Ask Before Choosing a Website Chatbot Vendor.

Frequently Asked Questions

Can an AI chatbot handle customer service outside business hours?

Yes, for the routine questions that make up most of the volume, like hours, directions, availability, and policies. Urgent or sensitive messages should be routed to a person, with the routing set up before launch.

Will an AI chatbot replace my customer service team?

No. It takes the repetitive, after-hours, and multilingual questions off their plate, and hands over the ones that need judgment. Your team spends less time on "what time do you close" and more on the conversations that need a person.

Can an AI chatbot answer in other languages?

A well-built one answers in the language the customer uses. On our deployments, that has included English, Spanish, and Hungarian.

How much maintenance does a website chatbot need?

Plan for a monthly review. Someone needs to read the conversations, update the knowledge base when your business changes, and check that escalations are reaching the right people.

See what a chatbot would handle for you

We set up and manage CathyAI chatbots built on your own information, with escalation paths and ongoing monitoring built in.

Learn more at emdigital.ca/cathyai-chatbot.