The Table Booking Sitting in Your Inbox While the Lunch Rush is Screaming (And How to Fix It)

Draft for Amber to review.

If you run a restaurant, café, or busy nail salon, you know the scene.

It is 12:18. The phone is ringing. Someone is waiting to pay. A table needs clearing. The kitchen wants to know whether table seven has an allergy. Your phone is vibrating with booking enquiries, Instagram messages, and the customer who simply wrote, “Hi, do you have space for four tomorrow?”

You mean to reply.

Then lunch gets louder.

By 2:47, that message is still sitting there. By 4:00, the customer has booked somewhere else.

You did not fail at AI. You hired it badly.

You probably tried ChatGPT once, asked it to write something bland about your business, got a paragraph that sounded like a bank advert, and thought, “Meh. Not for me.”

Fair enough.

The problem was not that AI is useless. The problem was that nobody showed you how to give it a proper job.

The issue is not your team. It is the gap between a ringing phone and a realistic reply.

Your staff are not ignoring customers. They are serving the customers physically standing in front of them.

That is the right priority.

But the booking enquiry still matters. A four-person table, a Saturday appointment, or a private dining request can be worth a lot to a small business. When the first reply comes hours later, the opportunity may already be gone.

The same thing happens in a nail salon. A client sends a message asking about an infill next week. The salon is full of people with their hands under lamps and foils. Nobody can answer properly without stopping what they are doing.

So the inbox quietly becomes a graveyard.

You do not need AI to run the whole business. You need it to handle the first reply.

This is where most AI advice gets silly.

You are told to automate everything. Replace every process. Build complicated systems. Connect seventeen tools. Then someone tries to flog you a subscription before asking how you currently take bookings.

No faff, please.

For a small business, the useful first step is much simpler:

1. Notice a new enquiry.
2. Reply quickly.
3. Ask for the important details.
4. Give accurate information.
5. Pass anything complicated to a human.
6. Record what happened.

That is it.

The AI does not need to pretend to be your restaurant manager. It does not need to make promises about allergens, squeeze in bookings you cannot fulfil, or invent a cancellation policy.

It needs a clear job description.

For example:

“You are the first-response booking assistant for [business name]. Reply warmly and briefly. Collect the customer’s preferred date, time, party size, name, phone number, and any relevant notes. Only offer availability that has been confirmed. Do not guess. Do not make promises about allergies, special requests, prices, or large parties. If the request is unclear or sensitive, tell the customer a member of the team will follow up.”

That is already more useful than asking an AI tool to “manage my bookings”.

Give your AI employee boundaries, not just access.

This matters because AI can be fast and still be wrong.

You need to decide what it can handle without you.

A sensible first version might reply to:

• Standard table booking enquiries.
• Opening hours.
• Location and parking information.
• Basic menu questions.
• Appointment availability, if the diary is connected.
• Simple booking changes or cancellations.
• Requests that arrive through email, web chat, SMS, or social messages.

It should send the following to a person:

• Large group bookings.
• Private dining or events.
• Allergy questions that require kitchen confirmation.
• Complaints.
• Refunds or payment issues.
• Anything involving a customer who sounds upset.
• Any request where the information is not in your approved notes.

This is not a weakness.

It is good workforce management.

You would not give a new administrator access to every system on their first morning, then disappear into the kitchen. You would give them a job description, show them the rules, and check their work.

AI staff need the same treatment.

That is the thinking behind Back of House by Phare HR, AI staff, set up and managed. We use the same practical principles that apply to people: clear responsibilities, proper onboarding, supervision, and regular updates.

Make the reply sound like you, not like a customer service department.

A corporate chatbot says:

“Thank you for your enquiry. We appreciate your interest and will respond at our earliest convenience.”

Nobody talks like that in a café.

Try this instead:

“Hi, thanks for getting in touch. We are serving lunch at the moment, but I can help you get the booking started. What date, time, and number of people were you thinking?”

That reply is warm. It is honest. It explains the delay without making the customer feel brushed off.

For a salon:

“Hi, thanks for messaging. We are with clients right now, so replies may be a little slower, but I can help you check the basics. What treatment are you after, and which day would suit you?”

Notice what it does not do.

It does not pretend a human is sitting there waiting. It does not use a daft amount of enthusiasm. It does not say “we value your business” twelve times.

It sounds like a real person who understands that the customer wants an answer.

You can teach this by giving the AI examples of how you already write. Include:

• A normal booking reply.
• A polite way to say no.
• Your preferred greeting.
• Words you use often.
• Words you would never use.
• How you explain delays.
• How you handle a customer who needs a human.

This is where Amplified Impact comes in. The aim is not to make your business sound more artificial. It is to help your existing voice do more work, without asking you to type the same reply during every lunch service.

Start with a first reply, not a fully autonomous booking machine.

You do not need to connect every system immediately.

In fact, starting small is often safer.

You could begin with an AI-assisted inbox that drafts the reply for you. During the lunch rush, it identifies the enquiry, gathers the missing details, and prepares a response. You quickly check it, edit if needed, and send.

That alone can remove a lot of the mental clutter.

The next version might send automatic replies to straightforward enquiries. The customer gets an immediate acknowledgement and answers a few simple questions. You receive a tidy summary when human input is needed.

Only after that would you consider letting the system check a live booking diary or create standard bookings automatically.

Why the caution?

Because the catches matter.

A booking system can be out of date. A staff member may have blocked a table informally. Your salon diary may not show that one technician is away. Your menu may have changed last week. Your AI will not magically know any of this unless somebody updates its information.

Speed is useful. Accurate speed is better.

So keep one simple source of truth. It might be your booking platform, shared calendar, or an agreed spreadsheet while you test the process. Make one person responsible for updating opening hours, holiday closures, treatment prices, menu changes, and booking rules.

If the information changes, the AI’s instructions need to change too.

Here is the method you can use today, for free.

Take ten minutes and write down the answers to these questions:

1. What enquiries arrive during your busiest hour?
2. Which ones are safe to answer automatically?
3. What details must be collected before a booking can be considered?
4. Which requests always need a human?
5. What should the first reply sound like?
6. Where should the conversation be recorded?
7. How quickly can someone on your team review anything that is escalated?

Then create three short documents:

Your business facts.
Opening hours, location, parking, booking link, services, menu basics, policies, and contact details.

Your booking rules.
Party-size limits, appointment lengths, deposit requirements, cancellation rules, blackout dates, and when to offer alternatives.

Your voice guide.
A few examples of how you write, plus a list of phrases that sound wrong for your business.

Give those documents to the AI tool you already use. Ask it to draft replies based only on that information. Tell it to say, “I need to check that with the team,” rather than guessing.

Then test it with real examples from last week’s inbox.

Not perfect examples. The awkward ones.

The customer who asked for a table “around seven-ish”. The person who wanted to bring a dog. The group of twelve asking for a last-minute booking. The client who sent only a screenshot and no explanation.

That is how you find the holes before your customers do.

The goal is not to remove the human touch. It is to protect it.

Your customers still want you.

They want your welcome, your judgement, your humour, and your ability to deal with the unusual situation that no script could predict.

They do not necessarily need you to type “Hi, what date were you thinking?” for the 46th time that week.

Let AI handle the first, repetitive layer. Let your team handle the moments where care and judgement matter.

That is a much more sensible use of the technology than trying to replace the whole front of house.

And if your first attempt at AI felt disappointing, you are not behind. You were probably given a tool without a role, a process, or any supervision.

That is not an AI problem. That is a management problem.

Luckily, management problems can be fixed.

If you want to see how this works in practice, we are covering it in the free class "You didn't fail at AI. You hired it badly." on Friday 25 September 2026, from 12:00 to 13:30 BST. You can register here: https://calendly.com/amberaziza/ai-for-your-business.