A customer inquiry is complex when it consists of multiple parts, lacks information, requires data from systems, or ultimately involves a decision. An AI chatbot solves the first three cases if it can ask follow-up questions, conduct dialogues, and access systems. For discretionary decisions, customer dissatisfaction, and knowledge not found in any source, it hands over to a human—along with the full conversation history. This article breaks complex inquiries down into their building blocks and outlines rules for effective handovers.
At a glance
Complexity stems from six typical causes: multiple sub-questions, missing information, system data, multi-step processes, discretion, and emotion.
A well-built chatbot resolves the first four on its own, using follow-up questions, guided dialogues, and integrations.
Discretion and emotion belong to humans.
If knowledge is missing, the chatbot acknowledges it and hands over.
A successful handover includes the conversation history and captured details so no one has to start from scratch.
What makes an inquiry complex
An example from technical service: "I need a replacement part for the dosing unit of our system, how long will delivery take, and is this still covered under warranty?" This single sentence contains three questions, one piece of missing information, and two system queries. That is precisely what makes it complex.
Multiple sub-questions. The customer asks about the part, delivery time, and warranty all in one sentence. A chatbot must recognize each component and answer them individually.
Missing information. Which system, which model series, which year of manufacture? Without these details, there is no correct answer. The chatbot asks targeted follow-up questions.
Data from systems. Delivery times and warranty status reside in the ERP or CRM. Without an integration, the chatbot cannot access them.
Multi-step processes. A complaint requires an order number, error description, and contact details—collected in a specific order and validated. This requires a structured dialogue.
Discretion. Goodwill gestures, special terms, an exception to the rule: this involves a decision that the company deliberately leaves to humans.
Emotion. An upset customer calling for the third time about the same issue needs someone who listens and takes action.
What an AI chatbot can resolve autonomously
Recognize sub-questions and answer them individually
A chatbot that only identifies the most likely intent will answer just one out of three questions. A better approach is intent recognition that can select and combine the appropriate information pathways: a knowledge base answer for warranty terms, a system query for delivery time, and product selection with follow-up questions for the spare part.
Ask targeted follow-up questions
When information is missing, asking a follow-up question is the right response. The chatbot inquires about the model series or serial number and proceeds from there, without requiring the customer to repeat their question. This also applies to ambiguous phrasing—for instance, when a customer describes a symptom but means a spare part.
Access back-end systems
With integrations into ERP, CRM, or ticketing systems, the chatbot can retrieve delivery times, order statuses, or warranty details directly within the ongoing conversation. Without these integrations, the only option left is routing. Why this represents the actual bottleneck for many LLM chatbots is discussed in the article Why LLM-based chatbots cannot resolve service cases without integration.
Guide multi-step processes
Structured dialogues are ideal for processes with fixed steps: they collect the required details in the correct sequence, validate them, and prompt with follow-up questions if inputs are invalid. Generative responses are used where open formulation is appropriate, while structured steps are applied where process compliance is essential.
When the chatbot should hand over
Four rules cover most scenarios.
The knowledge is missing. If the answer is not found in an approved source, the chatbot says so openly and hands over. A plausible yet fabricated answer costs far more in customer service than any escalation.
Discretion is required. Goodwill, price discounts, exceptions: these decisions are made by a human.
The customer is dissatisfied or asks for a human. In this case, the chatbot hands over immediately, without further automated attempts.
The process stalls. If a detail remains unclear after several follow-up prompts, a handover is faster than another loop.
Which topics a chatbot should avoid answering from the outset is covered in the article What a chatbot should not answer.
What a good handover looks like
A handover is only as good as the context it carries. The human agent taking over should see the conversation history, already captured details such as the system model, order number, or error description, and the reason for the handover. This allows the human conversation to start directly with the actual issue.
Routing is equally important: an inquiry about a spare part belongs with technical service, an invoice question goes to accounting, and an English inquiry to an English speaker. Once resolved, the conversation can be routed back to the chatbot, for example, to provide a confirmation or handle the next step.
How Mercury.ai processes complex inquiries
At Mercury.ai, a model orchestration engine identifies the user's intent and selects the optimal information pathway: a knowledge response with follow-up questions, a product selection flow, or a structured dialogue. If information is missing or the facts are ambiguous, the chatbot asks clarifying questions or transitions to a structured dialogue. Through integrations, it reads from and writes to CRM, ERP, and ticketing systems during the active conversation. This significantly reduces hallucinations, as answers originate from approved sources and are validated before output.
If a human is needed, the chatbot transfers the case within the Agent Desk complete with the full conversation history and pre-qualified data—routed by department, language, or workload capacity. After the live chat, the dialogue returns to a predefined point in the chatbot flow. Learn more about the underlying architecture on our Mercury Intelligence page. Discover which recurring tasks can be fully automated in our article on 15 use cases in customer service.
Conclusion
Complex inquiries are typically compound inquiries. A chatbot capable of asking follow-up questions, conducting guided dialogues, and accessing systems can successfully resolve multiple sub-questions, missing data, system lookups, and multi-step processes. Discretion, emotion, and missing knowledge belong to human agents—and handing over to them is an integral part of an effective solution. Clearly separating the two creates a chatbot that customers will trust even with difficult questions.
Want to evaluate which complex inquiries your chatbot can handle? Talk to us.
About the Author
Dr. Maximilian Panzner
Chief Technology Officer @Mercury.ai
Dr. Maximilian Panzner is CTO and co-founder of Mercury.ai. He earned his doctorate in computer science at the CITEC institute of Bielefeld University, where he conducted research on multimodal machine learning and intelligent interaction systems. For over 20 years, he has been working on artificial intelligence, human-machine interaction, and conversational AI platforms for enterprise use.
