Will customers be annoyed by an AI bot handling enquiries?

They will if the automation impersonates a human, cannot answer and will not let them through. Those three things must be ruled out: transparency from the first message, one step to reach a person, and handing the case on when a question exceeds the assistant's knowledge. The alternative to an automation is usually not an immediate human conversation but waiting until the next day.

Where does an AI assistant get answers to customer questions?

Only from the company's own material: documents, price lists, terms and data from systems it is given access to, not from the model's general knowledge. Answer quality therefore depends on document quality — if terms are described in several places and each differs, that must be settled first.

Does automating customer enquiries mean cutting the support team?

That depends on what the company does with the reclaimed time and is an owner's decision rather than an effect of technology. More often it means handling more cases with the same team. The work that disappears is answering the same question repeatedly; what remains is harder and needs different support.

CUSTOMER ENQUIRIES — TRIAGE · ANSWER · HAND OVER

CUSTOMER ENQUIRIES

Repetitive questions about status and terms go to the automation. Hard cases stay with people — and the customer gets an answer before they think to phone.

Free consultation See the split

WHAT AI TAKES, AND WHAT STAYS WITH A HUMAN

Customer enquiries are the easiest area to get wrong with AI — wrong meaning the customer feels they are talking to a machine guarding the door to your company. So we start by separating what the automation touches from what it doesn't.

AI
takes REPETITIVE, VERIFIABLE, NO JUDGEMENT REQUIRED

Order, delivery and case status — answered straight from the system
Questions about availability, pricing, terms, documents to download
Initial triage: what the case is about and who it should reach
Gathering details before a human picks the case up
Answers outside working hours, when the alternative is silence until morning

A human
keeps WHEN JUDGEMENT, MONEY OR EMOTION IS INVOLVED

Complaints and anything the customer experiences as a dispute
Negotiating price, deadline or scope
Unusual cases nobody has seen before
Every instance where the customer asks for a person
Decisions somebody has to take responsibility for

FOUR RULES WE DON'T BREAK

01 · TRANSPARENCY

The customer knows it's an automation

We don't impersonate a human and we don't give the assistant a name designed to suggest otherwise. Pretending works until the first slip, and then costs more trust than the automation ever saved in time.

02 · ESCAPE HATCH

One step to reach a person

Always visible and always working, with no menu to navigate and no need to ask three times. An automation you can't get out of generates more phone calls than it answers messages.

03 · LIMITS OF KNOWLEDGE

„I don't know” instead of inventing

The assistant answers only from your documents and data from your systems. When a question goes beyond that, it hands the case on rather than composing something that merely sounds plausible.

04 · CONTENT CONTROL

The knowledge is yours and editable

Answers come from documents your team can correct without calling us. Changing a price list or delivery terms is editing a file, not raising a ticket and waiting a week.

WHAT TO EXPECT FROM THE NUMBERS

In a typical 20–200-person company, repetitive questions about status and terms make up most of the incoming messages. Handing those to an automation doesn't eliminate the support team — it means the same people work on cases that genuinely need a conversation, and the customer gets an answer sooner.

Customer enquiries are one of four areas where we deliver AI — the rest are on AI delivery. If the assistant is to know statuses, it needs access to data, so system integration usually appears along the way.

FREQUENTLY ASKED

Won't customers be annoyed at talking to a bot?

They will, if the automation impersonates a human, can't answer and won't let them through. Those are exactly the three things we rule out: transparency from the first message, one step to a human, and handing the case on when the question exceeds what the assistant knows.

Set up that way, irritation drops — because the alternative to the automation usually isn't an immediate conversation with a person, it's waiting until tomorrow.

Where does the assistant get its answers from?

Only from your material: documents, price lists, terms and data from the systems it is given access to. It doesn't answer from the model's general knowledge, because that contains nothing about your company.

The practical consequence is that answer quality depends on document quality. If delivery terms are described in three places and each says something different, that has to be settled first — and it is usually the first benefit of the project.

Does this mean cutting the support team?

That depends on what you do with the reclaimed time, and it is an owner's decision rather than an effect of the technology. In the companies we work with it more often means handling more cases with the same team than making redundancies.

The work that disappears is answering the same question for the hundredth time. The work that remains is harder and usually more interesting — but it also needs different support than before.

WE REPLY WITHIN 24 HOURS