Canonical: https://nexuri.de/en/guides/customer-requests Language: en Practical guide · Customer enquiries # Turn a customer enquiry into a clear workflow. Automation helps where information is repeatedly read, copied and passed between people. This guide outlines a possible starting point using AI, existing systems and clear human approval. [Discuss your workflow](https://nexuri.de/en/ai-check) [Explore the example](https://nexuri.de/en/guides/customer-requests#example) 01 Start with a recurring request 02 Define approvals and errors first 03 Evaluate practical benefit in a pilot 01 / A worked example ## An email arrives. Context follows. An illustrative workflow for a facility management team. This example explains the approach; it is not a documented customer result. Example incoming message > “The stairwell lighting at property A has not worked since yesterday. Could you arrange a repair?” - 01 Capture. The message is recorded once as a task. Sender, property and issue are extracted; missing details remain visible. - 02 Add context. Approved property information and existing tickets are checked. A possible duplicate is flagged rather than creating another job. - 03 Review the draft. A ticket and reply are prepared. The responsible person checks urgency, ownership and next steps. - 04 Hand over & record. After approval, the task is updated in the intended system. Sending, status changes and decisions are recorded traceably. Signs of danger or missing information require a direct handover to your team. Emergencies follow your established emergency procedures. 02 / Systems & data ## Understand the path the information takes. An initial pilot does not require an entirely new platform. The important questions are which systems are involved and where reliable information lives. In short A pilot for automating customer enquiries needs one defined entry point, an authoritative data source for each piece of information and minimal permissions. A new platform is rarely required – what matters is connecting existing systems cleanly through available integrations. ### One clear entry point. For example, an approved shared inbox or form. Define which requests are included in the pilot and which are outside its scope. ### An authoritative data source. CRM, property management or ticketing: each data point should have a clear source of truth and an available integration. ### Only the necessary permissions. Define which data can be read and which records can be changed. Test accounts and sample data help validate the workflow before operation. 03 / Decisions & exceptions ## Prepare automatically. Approve deliberately. Rules suit clear tasks. AI can structure free text and create drafts. Approval boundaries determine when your team takes over. In short AI prepares the ticket and the reply; a person approves anything binding. Commitments, costs and appointments are reviewed, errors land in a visible queue, and repeated messages are recognised as duplicates instead of creating a second job. ### Approval before commitments. Promises, spending decisions, appointments and sensitive replies should receive appropriate human review within the agreed process. ### Make errors visible. A failed integration, conflicting details or an unknown property must not carry on unnoticed. The task goes into a queue that can be reviewed. ### Avoid duplicate actions. Define how a previously handled request is identified and what happens on a retry. A repeated message should not simply create another job. 04 / The pilot ## Measure more than speed. Document the current process and compare it with a limited test. That turns an appealing example into an informed decision. In short A pilot compares today’s process with a limited test. It measures active handling time including review, the quality of classification and drafts, and ongoing cost and ownership. Criteria for continuing, adjusting or stopping are agreed before it starts. ### Handling effort. How much active work does a comparable request require? Include review time and later corrections. ### Quality & handovers. Are classification and drafts correct? How many follow-up questions, wrong assignments or manual interventions occur? ### Costs & responsibility. Include usage costs, maintenance and operational ownership. Agree when the pilot should continue, change or stop. Key terms ## The vocabulary for your pilot. Human-in-the-loopA workflow in which a person reviews or approves certain results before they become binding – such as a reply to a customer or an instruction to a contractor. Authoritative data sourceThe system that is the source of truth for a given piece of information, such as the CRM for contact details or property management for buildings. Other systems take that information rather than overwrite it. IdempotencyRunning an action more than once produces the same result. In a workflow, a message delivered twice does not create a second job. Exception queueA reviewable place for tasks that could not be completed automatically. Your team sees them there instead of them disappearing unnoticed. For your first conversation ## Four things make a useful starting point. No lengthy presentation needed. Bring one concrete task. - A typical request without unnecessary personal data - The systems you currently use - The current path from receipt to response - The step that takes most of your team’s time From our work ## How AGS connected its workflows. Our existing AGS case study covers ticketing, scheduling, contractor coordination and reporting. It provides an actual project alongside this guide’s illustrative example. [Read the AGS case study](https://nexuri.de/en/work/ags-facility-management) Common questions ## Define the right starting point. Should every reply be sent automatically? That is usually not a useful starting point for an initial pilot. Begin with drafts and a clear approval step. Which routine replies might later be sent automatically depends on content, risk and test results. What if AI misunderstands information? The process needs reviewable drafts, visible missing details and a defined human handover. Test cases should include spelling mistakes, incomplete messages and conflicting information. When is a simple rule better than AI? When a decision can be described precisely, such as routing a known form type to a fixed team. AI becomes more relevant where natural language needs to be interpreted or information summarised. Useful next steps ## See the workflow in your context. [### For facility management teams Requests, coordination and reporting as one connected workflow.](https://nexuri.de/en/solutions/facility-management) [### For real estate teams Connect enquiries with property knowledge and your CRM.](https://nexuri.de/en/solutions/real-estate) [### AI agents & MCP Agents that handle tasks in your CRM and ticketing with approval.](https://nexuri.de/en/services/ai-agents) [### Intelligent document processing Read and validate attachments such as invoices and forms automatically.](https://nexuri.de/en/services/document-processing) From the example to your team ## Which task would you like to handle fewer times? In a free AI opportunity check, we discuss your workflow and look for practical, realistic opportunities to automate it. [Request a free AI check](https://nexuri.de/en/ai-check) A no-obligation initial conversation. Implementation and scope are agreed separately.