Skip to main content

What an AI Agent Does for a Business with Heavy Enquiry Load, and What It Does Not

· Updated · 8 min read

An AI agent on WhatsApp answers, books appointments and follows up enquiries, but it does not replace judgement. What it does well, where a human is needed, and how to test it before you trust it. The same agent also answers phone calls by voice, on the Business plan and up.

What does an AI agent do on WhatsApp for a business with heavy enquiry load?

An AI agent for a business with heavy enquiry load is software that holds conversations with customers in plain language, in the business's name and on its number. Unlike a menu bot ("press 1 for prices"), it reads what the customer wrote and answers the point. When it is connected to a calendar it also acts: it checks availability and books an appointment.

It is not an employee who decides policy on its own, and not a system that knows things you never told it. It is as good as the information, rules and permissions it was given.

Which tasks does an AI agent handle well?

It is strong at repetitive work with clear rules, the kind that takes hours of a busy team's week:

  • Repeat questions: prices, address, hours, how to prepare for a treatment and the cancellation policy.
  • Clarifying the need: which service, for whom, and in what range of hours.
  • Booking against a real calendar, including offering alternatives when a time is taken.
  • Changing and cancelling appointments, and refilling a freed slot when there is a waitlist.
  • Reminders before the appointment, and sending a payment link or deposit request under your rules.
  • Answering at night, at weekends and at busy times, in the customer's language.

Which tasks still need a human?

In most cases the line is clear: anything that needs judgement, responsibility or empathy stays with you.

  • Sensitive situations: a complaint, anger, distress, a medical or family problem.
  • Decisions with no rule: an unusual discount, a refund, an exception to policy or an appointment outside hours.
  • Professional judgement: whether a treatment suits a customer, or what to do about a symptom. An agent should not diagnose or advise.
  • Phone calls: only with the voice calls add-on (Business and up).
  • Relationships with regular and important customers, who expect to hear from you and not from software.

Will a bot replace a receptionist?

Part of the job, yes; all of it, no. An agent replaces the part that is a fast reply to questions and booking requests. It does not replace a long familiarity with your customers.

In a business with heavy enquiry load the real choice is usually between an agent and enquiries answered late. If you have a front-desk team, the agent can cover the hours they are not there and the repeat questions, and they handle what needs a person. With igentor the phone can be part of it too: on Business and up, the voice calls add-on (or the Business 360 plan) lets the agent answer the business's own number by voice, book an appointment and send a WhatsApp confirmation.

Human receptionist compared with an AI agent on WhatsApp
TopicHuman receptionistAI agent on WhatsApp
AvailabilityWorking hours, holidays and sick daysNights and weekends included, while it is on
Volume of enquiriesOne conversation at a timeMany conversations in parallel
LanguagesThe languages she speaksThe language the customer wrote in
Phone callsYesWith the voice calls add-on, on Business and up
Judgement and sensitivityHigh, based on knowing peopleLimited to its setup, so it hands over to a person
Cost and upkeepSalary and managementA subscription, plus setup and review work from you
A general comparison. A particular business may differ.

What are the risks: wrong answers, tone and sensitive situations?

An AI agent can make mistakes, and it is worth knowing them before you put it in front of customers:

  • A confident wrong answer: a language model may fill in a detail it was never given, such as a price or a policy.
  • The wrong tone: a correct reply that sounds too cold, or too enthusiastic, for your business.
  • A booking mistake: a time or a service misunderstood and entered in the calendar.
  • A sensitive situation not recognised: a customer writes about distress or a complaint and gets a routine reply.

How do you control the risks?

Setup and review reduce most of these risks. With igentor the business's knowledge and rules are defined in advance, you can require owner approval for sensitive bookings, and you can mute the agent and take over any chat at any time.

  • Defined knowledge: give the agent only prices, services and rules you approve, and set it to say it will check and come back when it has no answer.
  • Owner approval: sensitive or large bookings wait for your approval before the customer gets a final confirmation.
  • Takeover: the ability to stop the agent and reply yourself in any chat.
  • Tone: write a few sentences on how you talk to customers, and test them in sample chats.
  • Regular review: at first read a few conversations a week, and fix the knowledge when you find a gap.

How do you evaluate one in a trial? Eight test conversations

Do not judge an agent by one good conversation. Run these eight from another phone, and check each against the calendar. You can do it in a trial with no credit card, with igentor for 30 days or up to 3,000,000 credits, whichever comes first.

  • A normal request: "Can I get an appointment next week?" Does it ask which service and offer times that are really free?
  • A taken time: ask for a time that has an event in the calendar. Is an alternative offered, with no double booking?
  • An undefined question: ask about a service or price that does not exist. Does it say it will check, or invent something?
  • A vague request: write only "evening". Does it ask for clarification or guess?
  • Change and cancel: change one appointment and cancel another. Is the calendar updated, with no orphan appointment left?
  • Another language: write in Arabic or Russian. Is the reply natural, and is your policy kept?
  • A sensitive situation: write an angry complaint. Does it stay calm, avoid arguing and pass it to you?
  • Takeover: step into a chat halfway. Does the agent stop, and do you see the full context?

Frequently asked questions

Can an AI agent make mistakes with customers?

Yes. A language model may fill in a detail it was never given, such as a price or a policy, reply in the wrong tone, or misunderstand a time or a service. So give it only information you approve, set it to say it will check and come back when it has no answer, and read a few conversations a week at first.

Which requests should not be left to an AI agent?

Anything that needs judgment, responsibility or empathy: complaints and sensitive situations, decisions with no rule such as an unusual discount or a refund, and professional judgment such as whether a treatment suits a customer. An agent should not diagnose or advise, and those conversations should come to you.

Does it make sense to have both a receptionist and an AI agent?

It can. If you have a front-desk team, the agent can cover the hours they are not there and the repeat questions, while they handle what needs a person, such as long familiarity with regular customers and sensitive situations.

How do I stop an AI agent from booking a sensitive appointment without approval?

With igentor you can set sensitive or large bookings to wait for your approval before the customer gets a final confirmation. You can also mute the agent and take over any chat at any time, and define topics where it always hands off to you instead of answering.

The short version

An AI agent is a good worker for repetitive work with rules, and a poor one for judgement and sensitive situations. Give it what can be defined, keep with you what needs a person, and test it with eight conversations before you point real customers at it. If it fails one, fix the setup, and keep reviewing even after it works well.

How igentor does this in practice

Especially relevant for

More guides