Use AI to Reply to Your Restaurant's Guest Reviews

What AI is useful for here

When the review screen fills up during the day, every line starts to look equally important. But a cold dish, slow service or a mixed-up delivery each needs a different response. Jev sorts the text into food quality, speed of service or delivery, answers yes or no to whether the manager should see it, and adds a numeric priority score when needed. Jev does not write the reply.

When only the lines that really need attention reach the manager, the conversation starts calmer. First you get clear on what slipped: a kitchen mistake, slow service, or a note on a delivery order. Then only that line goes into the chat prompt, and the reply draft keeps the name of the dish.

Two checks matter before you publish. The dish named in the review stays in the draft, because the guest wants to see that their own plate was noticed, not read a generic apology. Any one-size-fits-all apology gets deleted and replaced with what the business is actually going to check.

When the complaint flow is kept alongside the customer complaint form, coming back to the same issue the next day is much easier. On the form, the manager sees which dish, which shift and which channel the complaint was about.

This approach sits close to servant leadership in practice. The aim is not to defend yourself but to make the problem the guest pointed to clear enough for the team to see.

What information to give it

Before you give any text to Ritapos Ai, decide which pieces count as part of your business records. A short but complete input keeps the AI's answer more reliable.

• The channel the review came from and the date range, for example a delivery platform, QR menu feedback or a social media message.
• The guest's original text, typos included, without adding any names.
• Details from the review the business can check, such as the dish name, portion, sauce, how it was cooked or a note on the delivery order.
• Whether the order belonged to dine-in, takeaway or delivery, without setting up a separate tagging system on this page.
• What is already known from the shift, for example a kitchen delay, a busy service, a missing item or how long the courier waited.
• The threshold that means the manager must see it: a possible food safety issue, repeated delays, the same item complained about twice, or a harsh tone.
• The language of the platform where the reply will be posted and the business's tone of voice, for example calm, short and focused on a fix.
• Stock phrases to delete before publishing, for example the same generic apology everyone gets and explanations that avoid taking ownership.

AI prompts

First pass with Jev

This prompt sorts the incoming text by what the business needs to do about it. The manager does not see every review, only the lines that truly need them.

Role: Act as a calm classifier who reads reviews for a restaurant's operations team.
Facts given: Read the line [review text]. The category can only be food quality, speed of service or delivery. Answer yes or no to whether the manager should see it, and add a priority score from 1 to 5.
Do not: Do not write a reply draft, do not address the guest, do not invent names, do not produce star ratings, do not add explanations, do not promise rewards.
Output format: Category, manager should see it yes/no, and a priority score from 1 to 5.

Pick the lines for the manager

Out of a crowded list, only the lines the manager needs to look at are set aside. The team stops rushing to every review in the same panic.

Role: Act as a manager going through the complaint log at the end of a shift.
Facts given: From [Jev result], take the lines marked yes with a high score. Put repeat items, repeat problems or a harsh tone at the top.
Do not: Do not write explanations for lines marked no, do not produce automatic apologies, do not decide on compensation, do not add guest names.
Output format: List the lines for the manager with these fields: why it needs to be seen today, which record form it goes into, and who to ask.

A draft that keeps the dish name

The reply draft does not stay generic. Whatever dish the guest mentioned stays in the draft, and the business owns the problem.

Role: Act as an editor writing a short, calm, fix-focused reply on behalf of the owner.
Facts given: Use the dish name from [review sent to manager] exactly as written. State the [business check] in one sentence.
Do not: Do not fake warmth, do not describe an experience the guest did not have, do not promise a discount or a gift, do not drop the dish name.
Output format: Give a two-sentence reply draft. The first sentence names the dish, the second says what will be checked.

Strip out the generic apology

Before publishing, the text stops being an empty apology that could go to anyone. What the business is looking into becomes clear.

Role: Act as the final reader who trims a review reply before it is published.
Facts given: In [reply draft], keep the dish name, the thing to be checked and the calm tone.
Do not: Do not add new promises, do not write a sharp defense, do not reinterpret the review, do not leave the generic apology in.
Output format: Give the sentences to delete and the clean two-sentence draft that remains, separately.

Follow-up note for the form

The reply and the business follow-up are kept apart. The next day, the manager can see clearly what to check.

Role: Act as a service manager writing a short internal note for the customer complaint form.
Facts given: Write a form note based on [manager list] and [reply draft]. The topic must be food, speed of service or delivery.
Do not: Do not address the guest, do not write an unresolved issue as resolved, do not suggest automatic rewards, do not blame individuals.
Output format: Write four lines: topic, record to check, team to ask, and what to look at during close.

What to do when the answer runs long

This is not a page for writing reviews. When a business produces fake praise or invented experiences, it grows its own loss of trust. Here Ritapos Ai helps sort the incoming text, find the line the manager needs to see and prepare a short draft tied to the real issue.

Jev's job is classification. No labels are opened beyond food quality, speed of service and delivery. The question of whether the manager should see it stays a simple yes or no, with only a numeric priority score added when needed. Writing the reply happens in the chat prompt, and only after the line for the manager has been picked.

The dish name has to stay in the draft. If the guest said the soup was cold, the meat was tough or a garnish was missing, describing it as general dissatisfaction does not build trust. In the same way, ending every reply with the same apology makes it look like the business did not read the review.

This guide does not set up order notes or reservation tags. Those belong to other workflows. The focus here is the review's category, what the manager sees, the follow-up form and the short reply that gets published.

Use it with control forms and leadership

The customer complaint form is the business memory behind every reply. Once the reply is live, the team checks the form to see whether the same dish keeps coming up.

A servant leadership approach keeps the review from turning into a fight over who is right and focuses the team on what to fix. That is why it matters that the manager sees the line.

Sample output

The figures are examples. You fill in the form or the report yourself.

• Hello, we read your note about the [dish name].
• We will check [issue to check] in today's service records together with our manager.
• Your feedback will not stay a general comment. It will go into the notes for that shift.
• Thank you for sharing it with us.

A short reply is not always the easy one. For a short reply to be good, the dish name, the thing to check and a calm tone all have to be there together.

The fewer lines the manager has to look at, the better the follow-up. Treating every review as urgent makes the truly urgent ones look ordinary.

AI is not a referee making decisions for the business here. Ritapos Ai brings the right line to the table, and the final check still belongs to you.

Imagine three short reviews arrive after lunch service. One says the coffee was great, one asks about delivery time, and one says the [dish name] arrived cold. When Jev pulls out only the third line, the one that needs action, with its category, a manager yes and a score, the team does not blow praise up into a complaint and does not miss the real problem.