Filtering First Enquiries: AI Chat for Services

For agencies, consultants and service providers, the first enquiry is the moment that decides whether interest turns into a project. Yet not every message in the inbox is a project: alongside serious project enquiries there are quick factual questions, sales and partnership pitches, and requests that do not fit your field at all. Sorting all of that by hand costs time, and the genuinely good enquiries wait longer than they should. This is exactly where an AI chat assistant helps: it captures the first contact around the clock, asks the few questions that matter, maps the request to the right service and forwards only the pre-qualified enquiries to your team. That companies use AI above all in customer contact is shown by a recent survey: 88 percent (Bitkom) of firms that use AI apply it there. At the same time, customers want a reliable way to reach a human for real problems. This article shows how an assistant filters enquiries without scaring prospects off, and how a flooded inbox turns into an orderly list of fitting leads.

Enlarge graphic

Labels in the illustration

• Website relaunch
• budget approx. 8,000 EUR
• Qualified
• Opening hours?
• quick factual question
• Answered
• SEO link exchange
• cold sales pitch
• Off-topic
• Initial consultation
• wants an appointment
• Appointment
• XICBOT
• filters · qualifies
• Off-topic politely declined
• Reaches the team
• Qualified lead
• Request: website relaunch
• Budget: approx. 8,000 EUR
• Service: web development
• Slot: Tue 2 pm proposed
• Pre-qualified for the team
• Fewer unsuitable enquiries in the inbox
• Why it matters
• 7x
• higher chance to qualify a
• lead within 1 hour (HBR)
• 80%
• of common service issues
• resolved automatically by 2029 (Gartner)
• 62%
• of online shoppers prefer a
• human for problems (Bitkom)
• First contact around the clock, pre-qualified to the team

Filter enquiries, don't flood the inbox

Why Unsuitable Enquiries Cost Time

An inbox without a pre-filter treats every message the same. The urgent project enquiry sits next to the third question about opening hours and next to an offer for a link exchange that no one asked for. For a service provider already juggling client appointments, quotes and the actual work, that means double friction: first you have to sort out what is even relevant, and only then can you deal with it. The result is slow responses precisely where speed matters. Standard chatbots often do not solve this problem; on the contrary, they frequently reach their limits with complex questions, and anyone who then cannot find a way to a human is left alone with their request. A good assistant works the other way round: it answers the simple things itself and passes on the important ones in an orderly way.

Filtering here does not mean turning people away, but sorting them. The assistant recognises whether a message is a quick factual question, a serious project enquiry or something that does not fit the offer at all. The factual question it answers directly from your own content. The project enquiry it qualifies with a few follow-up questions and hands over in structured form. And the unsuitable request, such as a pure sales pitch, it deals with politely, without you even having to see it. So the team mainly receives what is worth the effort. How an individually trained assistant differs from an off-the-shelf chatbot is described in the article on a custom-trained assistant.

Capture the First Contact Around the Clock

Enquiries rarely arrive during office hours and neatly numbered. They come in the evening, at the weekend, during the lunch break, often at the moment of an acute need. Those who do not respond then frequently lose the prospect to the next provider who is faster. How much speed weighs is shown by an often-cited study: companies that contact a prospect within an hour are around seven times (Harvard Business Review) more likely to have a qualifying conversation with a decision maker than those who wait even one hour longer. B2B prospects generally expect a reply within 24 hours, while the fastest providers answer in under an hour (Forrester).

An AI assistant closes exactly this gap. It is reachable around the clock, captures the first contact at the moment of interest and keeps the prospect in the conversation instead of putting them off until later. It does not lead with a rigid form but asks one question after another, and only those the case requires. In this way a clear picture of the enquiry emerges within the first minutes, and the prospect feels taken seriously even when the team is not at their desk. How this guided first contact translates into clean lead capture is shown in the article on qualifying leads via chat instead of a long form.

Pre-qualify Need and Budget

The real work of the assistant begins after the greeting. From a few easy-to-answer questions it assembles a complete picture of the enquiry: what it is about, how urgent it is, what scope the project has and what budget range the prospect has in mind. The budget question in particular feels delicate, but can be asked naturally in conversation, for example as a choice of rough ranges rather than an open number. This separates early those who are planning a concrete project from those merely browsing. According to Gartner, agentic AI will resolve around 80 percent (Gartner) of common service issues without human intervention by 2029, reducing operational costs by about 30 percent (Gartner), which suggests how much routine could be automated in future, so that more time remains for the demanding cases.

• The request: which service or which problem is at the centre
• The urgency: whether it is pressing or can be planned calmly
• The scope: roughly how large the project is set up to be
• The budget range: broadly as a span, not a fixed figure
• The fit: whether the request falls within your field at all
• The contact: through which channel the prospect can be reached

Map to the Right Service

A service provider rarely offers just one thing. An agency distinguishes between a relaunch, ongoing support and individual campaigns; a consultant between an initial consultation, a project and a mandate. For the enquiry to reach the right contact, the assistant has to do more than qualify; it also has to map. Because it is bound to your own content, it knows your range of services and maps a request to the fitting service instead of giving a generic answer. What carries this mapping is well-kept, understandable content, as described on the knowledge base page.

This binding to your own sources is the decisive difference from a generic chatbot. The assistant is instructed not to invent services that do not exist and not to promise anything the team cannot deliver. If it recognises that a request lies outside the offer, it says so politely instead of producing a useless lead. And from the guided conversations the company learns along the way: which services are especially in demand, where prospects hesitate and which questions recur is made visible by conversation analytics.

From Conversation to Appointment

For many services the next logical step after the qualified enquiry is an appointment: an initial call, a consultation, an on-site visit. The assistant can initiate this step by asking for the preferred time frame, suggesting free times from the stored appointment types and taking an appointment request that you then confirm, instead of putting the prospect off until a later callback. In this way the arc closes from first interest to appointment request in one flow. How scheduling in chat works in detail is shown on the booking assistant page as well as in the article on appointment requests via chat assistant.

Whether appointment or classic enquiry, at the end there is a structured lead that arrives where it is processed. Instead of an informal message from which someone has to pick out the essentials, a clean record arrives, with the request, urgency, mapped service and a way to make contact. Through clearly permitted actions the assistant can hand this lead to the inbox or, in the Custom package, to a connected system, as described on the tool control page. The embedding on the website itself happens with a short code snippet, explained under integration.

Assistant, Form or Answering Machine?

An assistant does not replace every other route to an enquiry across the board; it fills a gap. A lean form is enough when enquiries are rare and identically structured. An answering machine records but qualifies nothing and answers nothing. And a human-staffed live chat only plays to its strength during the hours when someone is at their desk. The AI assistant sits in between: reachable around the clock, qualifying already in the conversation and handing over to a human as soon as it is needed. An indication of why the combination matters comes from satisfaction with customer service in online shopping: of those who had contact with a human agent there, around 86 percent (Bitkom) are satisfied, and of those who used a customer service chatbot, about 50 percent (Bitkom). An assistant that takes on routine and passes on the rest to humans combines both. The following comparison places the routes side by side.

Examples by Service

Which questions make sense depends on the service. A web agency separates, with a few follow-up questions, the planned relaunch with a fixed budget from the vague idea and the pure support request of an existing client. A management consultant clarifies the topic area, company size and time horizon before an initial call is scheduled. An IT provider or software vendor distinguishes between a trial enquiry, a concrete project and a support case and routes each strand to the right place. In all cases the same pattern applies: few questions, clear mapping, structured handover.

In more regulated fields the limits are drawn more tightly. A law firm may ask about the area of law, deadlines and jurisdiction, but gives no legal advice and remains a signpost. In real estate what counts above all is whether genuine interest exists, for which property and with which budget. And where a service provider also runs an online shop, the same conversational logic applies there too, for example when the assistant recovers abandoned carts in chat. An overview of the cases per sector is given on the industries page; the matching scope of functions for enquiries is bundled by the lead assistant.

Limits and an Honest Approach

Being honest means naming the limits. An assistant filters and qualifies; it does not decide. The actual assessment of an enquiry, the offer and the relationship with the client remain a matter for people. Customers see it similarly: when there is a problem with an online order, 62 percent (Bitkom) of online shoppers would like to turn to a quickly reachable human contact, and only 36 percent (Bitkom) want a chatbot for that. This is not a contradiction to using an assistant but its job description: it takes on the first contact and the routine so that time remains for the important conversations. Market forecasts also call for a sober view: Gartner expects more than 40 percent (Gartner) of agentic AI projects to be canceled by the end of 2027 due to escalating costs, unclear business value or inadequate risk controls.

The basis remains data protection, because even the first contact processes personal data. XICBOT stores this on servers in Germany, on the basis of a data processing agreement and with a clear deletion concept; the answers are generated by a specialised model provider, also outside the EU on the basis of standard contractual clauses, and the assistant is meant to ask only for what is needed to handle the enquiry. This data economy is at the same time a signal of trust, as the privacy and hosting page explains. Where a case is complex, sensitive or unclear, the assistant is meant to hand over to a human, with the full conversation as context, as the article on a clean handover to human agents shows. Which package suits your needs is set out on the pricing page. How an assistant for service providers is set up is shown on the industry page AI assistant for service providers; we are glad to discuss your own enquiry path in an initial conversation.

• Capture the first contact around the clock instead of putting prospects off
• Pre-qualify need, scope and budget with a few follow-up questions
• Map the request to the fitting service and the right contact
• Initiate appointments and close the arc to the arranged conversation
• Politely decline unsuitable enquiries instead of flooding the inbox
• Hand complex or sensitive cases to a human with full context
• Analyse conversations in a privacy-compliant way and keep improving the assistant

Frequent questions