In many IT teams, every new ticket starts with the same manual work: read it, classify it, check the knowledge base for whether the problem has come up before, and then hand it off to the right team. When ticket volume is high, this triage alone eats up noticeable time that is actually needed for the solution itself.
\nThe Ticket Pipeline takes over exactly this first step. A language model reads the ticket description, assigns the problem to one of four categories (hardware, software, network, access), and estimates its urgency. The workflow then searches the knowledge base for similar, already resolved cases and writes a structured solution proposal as an internal comment on the ticket. After that, the ticket is automatically assigned to the responsible team based on its category.
\nA human stays in the loop at the decisive point: the support team reviews the proposed solution and decides whether it fits before a reply goes out to the person who submitted the ticket. The workflow prepares the ground, it does not resolve tickets on its own and it does not reply directly to customers. For standard cases with good knowledge base coverage, this saves noticeable time; for unusual or new problems, at worst it only delivers the categorization, and the team takes over the rest as usual.
\nCopy the prompt below in full into your automation tool or agent builder. As a file: ticket-schiene.en.json
\n" } ], "schritte": [ { "nr": 1, "titel": "Trigger: new ticket", "beschreibung": "A new IT support ticket with a ticket ID, subject and description comes in and starts the process.", "rolle": "mensch" }, { "nr": 2, "titel": "Categorize", "beschreibung": "A language model assigns the problem to one of the four categories hardware, software, network or access, and estimates the urgency on a scale from 1 to 5.", "rolle": "automatisch" }, { "nr": 3, "titel": "Search the knowledge base", "beschreibung": "The workflow searches the connected knowledge base for similar, already resolved tickets and matching solution articles for the identified category.", "rolle": "automatisch" }, { "nr": 4, "titel": "Create a solution proposal", "beschreibung": "From the matches found, a structured comment is created with a problem summary, the three most likely solutions and links to the matching articles.", "rolle": "automatisch" }, { "nr": 5, "titel": "Assign the team", "beschreibung": "Based on the mapping of category to team, the ticket is automatically forwarded to the responsible unit.", "rolle": "ergebnis" } ], "herausgeber": "Voyage Digital", "version": "2.0", "stand": "2026-07-26", "nutzen": [ { "titel": "Quick first assessment", "text": "The solution proposal is available as a structured comment as soon as the knowledge base search is complete." }, { "titel": "Reliable routing", "text": "Tickets are automatically assigned to the responsible team based on the detected category." }, { "titel": "Knowledge base put to use", "text": "Already resolved, similar cases feed directly into the solution proposal." }, { "titel": "Lighter triage", "text": "Manual initial review and knowledge base search are no longer needed for standard cases." } ], "voraussetzungen": [ { "label": "Ticket system", "text": "Zendesk, Freshdesk, Jira Service Management or comparable, with API access and write permissions for comments and assignments.", "icon": "support_agent" }, { "label": "Knowledge base", "text": "Notion, Confluence, GitBook or another system with maintained solution articles, searchable via the respective API.", "icon": "menu_book" }, { "label": "Team mapping", "text": "A fixed assignment of category (hardware, software, network, access) to the responsible team.", "icon": "hub" }, { "label": "LLM access", "text": "API access to a language model, for example from OpenAI, Anthropic or Google.", "icon": "key" } ], "betrieb": { "testing": [ { "label": "Test ticket", "text": "Run through a realistic problem, for example a WiFi connection that does not work in the office.", "icon": "confirmation_number" }, { "label": "Categorization", "text": "Check whether the ticket is correctly assigned to one of the four categories.", "icon": "category" }, { "label": "Solution proposal and routing", "text": "Check whether the knowledge base matches are relevant and the right team is assigned.", "icon": "lightbulb" }, { "label": "Edge cases", "text": "Test the five most common IT cases and at least one case with no match in the knowledge base.", "icon": "rule" } ], "goLive": [ { "label": "Approval to start", "text": "Activate the workflow only once categorization, solution proposal and team routing are reliably correct.", "icon": "check_circle" }, { "label": "Pilot", "text": "Start with a single team or a single queue instead of going straight to full operation.", "icon": "playlist_add_check" }, { "label": "Monitoring", "text": "Track the rate of automatically resolved tickets and the escalation rate.", "icon": "monitoring" }, { "label": "Self-learning loop", "text": "Collect misclassifications and refine the prompt based on these cases.", "icon": "autorenew" } ], "haeufigeFehler": [ { "label": "Categorization is imprecise", "text": "Extend the prompt with examples for each category.", "icon": "tune" }, { "label": "No knowledge base matches", "text": "Check whether the search is too narrow or the knowledge base indexing is incomplete.", "icon": "search_off" }, { "label": "Incorrect team assignment", "text": "Validate the mapping of category to team.", "icon": "alt_route" }, { "label": "API authentication fails", "text": "Check tokens and write permissions.", "icon": "lock" } ] }, "arbeitsprompts": [ { "titel": "Ticket Categorization", "beschreibung": "Assigns an incoming ticket to one of the four categories and estimates its urgency.", "prompt": "You are an IT support assistant who analyzes and categorizes incoming support tickets.\n\nAnalyze the following support ticket:\n\nTicket ID: [TICKET_ID]\nSubject: [TICKET_SUBJECT]\nDescription: [TICKET_DESCRIPTION]\n\nYour task:\n1. Identify the main problem in the ticket description.\n2. Assign the problem to EXACTLY ONE of the following categories:\n- Hardware (physical devices, printers, monitors, keyboards and similar)\n- Software (applications, programs, licenses, updates and similar)\n- Network (WiFi, VPN, connection issues, internet outages and similar)\n- Access (login problems, passwords, permissions, account management and similar)\n3. Rate the urgency on a scale from 1 (low) to 5 (critical).\n\nRespond in the following JSON format:\n{\n \"category\": \"[Category]\",\n \"urgency\": [1-5],\n \"reasoning\": \"[Brief reasoning in one to two sentences]\"\n}" }, { "titel": "Solution Proposal", "beschreibung": "Creates a structured comment with solution suggestions from the knowledge base matches found.", "prompt": "You are an IT support expert who creates solution proposals for support tickets.\n\nCurrent ticket:\n- Category: [CATEGORY]\n- Subject: [TICKET_SUBJECT]\n- Description: [TICKET_DESCRIPTION]\n\nKnowledge base articles and similar tickets found:\n[KNOWLEDGE_BASE_RESULTS]\n\nYour task:\nCreate a structured comment for the support team with the following information:\n1. Summary of the problem (one to two sentences)\n2. The three most likely solution suggestions based on the articles found\n3. Links to the relevant knowledge base articles\n4. Recommended next steps for the support team\n\nFormat the comment professionally and practically. Use markdown for better readability.\n\nImportant: if no matching articles were found, provide general best practices for the [CATEGORY] category." } ], "prompt": "GOAL\nBuild a workflow that starts automatically on every newly incoming IT support ticket: categorization, a knowledge base search, deriving a solution proposal, and automatically assigning the ticket to the responsible team.\n\nAdapt the structure to the concepts of your automation tool (for example nodes, steps, zaps or agent actions). While doing so, preserve the trigger logic, the sequence and the data flow.\n\nTRIGGER\nType: webhook or native integration with your ticket system (Zendesk, Freshdesk, Jira Service Management or comparable).\nInput: ticket ID, subject, description.\n\nSTEPS\n1. AI step: categorization. Use exactly the Ticket Categorization working prompt (see the Working Prompts section). Output: JSON with category, urgency (1 to 5) and reasoning.\n2. Search the knowledge base: using the category and the core terms from the description, find similar, already resolved tickets and matching solution articles, including links to the matches.\n3. AI step: solution proposal. Use exactly the Solution Proposal working prompt (see the Working Prompts section) with the matches found as input. Output: formatted markdown comment.\n4. Add the comment: save the solution proposal as an internal comment on the ticket, not as a reply to the customer.\n5. Assignment: assign the ticket automatically via the category to team mapping (hardware, software, network, access). Optionally derive the priority from the urgency.\n\nPROMPTS\nTake over both working prompts from the section of the same name unchanged. Do not alter the placeholders they contain, and map them as described in the data flow.\n\nDATA FLOW\nTrigger to step 1: ticket ID, subject, description.\nStep 1 to steps 2, 3 and 5: category; urgency additionally for an optional priority.\nStep 2 to step 3: knowledge base matches found.\nStep 3 to step 4: formatted comment.\n\nERROR HANDLING\nNo knowledge base matches: do not abort the run. In this case, the Solution Proposal working prompt provides general best practices for the identified category.\nNo valid JSON from step 1: retry once, then assign the ticket to the standard triage team and log the error.\nCategory not clearly identifiable: choose the most likely category and state the uncertainty in the reasoning.\nAPI error while commenting or assigning: log the error and notify the workflow owner.\n\nPLACEHOLDER CONVENTION\nAll placeholders in square brackets inside the two working prompts are fully mapped dynamically, as described in the data flow. There are no placeholders that require manual configuration in the prompts themselves. Only the system access credentials and the category to team mapping are configured manually.\n\nSELF CHECK BEFORE COMPLETION\nCheck, before you report the workflow as finished:\n1. Are all five steps connected correctly and in the right order?\n2. Are both working prompts inserted in full and unchanged, and are all placeholders mapped?\n3. Is the JSON from step 1 parsed correctly, and is the comment saved as an internal comment, not as a reply to the customer?\n4. Are all points listed that require manual configuration: ticket system access, knowledge base access, the category to team mapping, and LLM access?", "einrichtung": { "intro": "Two routes lead to the same result. Pick the one that fits your setup.", "varianten": [ { "variante": "Builder-Prompt", "einleitung": "Suitable when you want to build the workflow in your own automation tool.", "schritte": [ "Open your workflow tool, for example n8n, Make, Zapier or Langdock.", "Create a new workflow and give it a name.", "Choose the AI-assisted build option if your tool offers one.", "Paste the builder prompt from above in full.", "Check the result against the step sequence in this document, then add your access credentials and channels." ] }, { "variante": "JSON-Import", "einleitung": "Suitable when you want to take the workflow over unchanged.", "schritte": [ "Download the JSON file from the JSON export section.", "Open the workflow area of your tool and create a new workflow.", "Give it a name and confirm.", "Open the menu on the workflow name and choose to import a JSON file.", "Upload the file, then add your own access credentials.", "Run a test before you activate the workflow." ] } ] }, "exportMeta": { "datei": "w09_ticket-schiene__workflow.json", "format": "voyage-workflow/1" }, "download": { "schema": "voyage-workflow/1", "name": "Die Ticket-Schiene", "beschreibung": "Kategorisiert ein neu eingehendes IT-Support-Ticket automatisch, durchsucht die Knowledge Base nach passenden Lösungen, erstellt einen strukturierten Lösungsvorschlag als internen Kommentar und weist das Ticket dem zuständigen Team zu.", "trigger": { "typ": "Webhook", "bedingung": "Das angebundene Ticket-System sendet bei jedem neuen Ticket einen POST-Request an die vom Workflow bereitgestellte Webhook-URL. Mindestens Ticket-ID und Beschreibung werden erwartet, Betreff, Priorität, Ersteller und Zeitstempel können optional mitgeschickt werden. Die Webhook-URL lässt sich zusätzlich mit einem Secret absichern." }, "schritte": [ { "nr": 1, "titel": "Trigger: neues Ticket", "rolle": "mensch", "aktion": "Das angebundene Ticket-System sendet bei jedem neuen IT-Support-Ticket einen Webhook-POST-Request an die vom Workflow bereitgestellte URL. Erwartet werden mindestens die Ticket-ID und die Problembeschreibung, Betreff, Priorität, Ersteller und Zeitstempel können je nach Ticket-System zusätzlich mitgeschickt werden.", "prompt_ref": null, "eingang": [], "ausgang": [ "Ticket-ID, Betreff, Beschreibung (Webhook-Payload des Ticket-Systems)" ] }, { "nr": 2, "titel": "Kategorisieren", "rolle": "automatisch", "aktion": "Ein KI-Agent liest Ticket-ID, Betreff und Beschreibung mit dem Arbeitsprompt Ticket-Kategorisierung und erzeugt in einem Durchgang drei strukturierte Felder: category (Hardware, Software, Netzwerk oder Zugang), team (Name des zuständigen Teams, abgeleitet aus den im Prompt hinterlegten Team-Zuständigkeiten) und summary (kurze Zusammenfassung des Problems). Die Teamzuordnung entsteht bereits in diesem Schritt und nicht erst am Ende des Ablaufs.", "prompt_ref": "arbeitsprompt-1", "eingang": [ "Ticket-ID, Betreff, Beschreibung" ], "ausgang": [ "category", "team", "summary" ] }, { "nr": 3, "titel": "Knowledge Base durchsuchen", "rolle": "automatisch", "aktion": "Eine Dateisuche durchsucht den vorab ausgewählten Knowledge Folder mit category und summary als Sucheingabe und liefert bis zu zehn relevante Dokumenten-Chunks zurück, jeweils mit Text, Dateiname, URL und Relevanz-Ranking. Die Trefferzahl lässt sich über das Feld maxResults zwischen 1 und 50 einstellen.", "prompt_ref": null, "eingang": [ "category", "summary", "Knowledge Folder (vorab manuell ausgewählt)" ], "ausgang": [ "Bis zu 10 Dokumenten-Chunks (Text, Dateiname, URL, Relevanz-Ranking)" ] }, { "nr": 4, "titel": "Lösungsvorschlag erstellen", "rolle": "automatisch", "aktion": "Ein zweiter KI-Agent liest mit dem Arbeitsprompt Lösungsvorschläge die gefundenen Dokumenten-Chunks zusammen mit category und summary und erzeugt drei strukturierte Felder: suggestions (formatierte Liste konkreter Lösungsschritte), confidence (hoch, mittel oder niedrig, abhängig von Anzahl und Qualität der Treffer) und related_tickets (Anzahl der gefundenen ähnlichen Fälle).", "prompt_ref": "arbeitsprompt-2", "eingang": [ "Bis zu 10 Dokumenten-Chunks (Text, Dateiname, URL, Relevanz-Ranking)", "category", "summary" ], "ausgang": [ "suggestions", "confidence", "related_tickets" ] }, { "nr": 5, "titel": "Team zuweisen", "rolle": "ergebnis", "aktion": "Ein Code-Schritt fasst die bisherigen Ergebnisse zu fünf Feldern zusammen: category, assignedTeam (übernommen aus team), comment (aus suggestions und confidence formatierter interner Kommentartext), confidence und ticketId. Diese Daten liegen damit strukturiert für das Ticket-System bereit. Das eigentliche Schreiben des Kommentars und die Neuzuweisung im Ticket-System selbst sind im Originalexport nicht enthalten und müssen als zusätzliche Aktion mit Schreibrechten ergänzt werden.", "prompt_ref": null, "eingang": [ "suggestions", "confidence", "related_tickets", "category", "team", "Ticket-ID" ], "ausgang": [ "category", "assignedTeam", "comment", "confidence", "ticketId" ] } ], "prompts": { "arbeitsprompt-1": { "titel": "Ticket-Kategorisierung", "text": "Sie sind ein IT-Support-Assistent, der eingehende Support-Tickets analysiert und kategorisiert. Analysieren Sie das Support-Ticket (Ticket-ID, Betreff, Beschreibung), identifizieren Sie das Hauptproblem, ordnen Sie es GENAU EINER der Kategorien Hardware, Software, Netzwerk oder Zugang zu und bewerten Sie die Dringlichkeit auf einer Skala von 1 bis 5. Antworten Sie als JSON mit kategorie, dringlichkeit und begründung." }, "arbeitsprompt-2": { "titel": "Lösungsvorschläge", "text": "Sie sind ein IT-Support-Experte, der Lösungsvorschläge für Support-Tickets erstellt. Erstellen Sie aus Kategorie, Ticket-Angaben und den gefundenen Knowledge-Base-Treffern einen strukturierten Kommentar mit Problemzusammenfassung, den drei naheliegendsten Lösungsvorschlägen, Links zu den relevanten Artikeln und empfohlenen nächsten Schritten. Formatieren Sie in Markdown. Ohne passende Treffer geben Sie allgemeine Best Practices für die Kategorie an." } }, "datenfluss": [ { "von": 1, "nach": 2, "nutzlast": "Ticket-ID, Betreff, Beschreibung aus dem Webhook" }, { "von": 2, "nach": 3, "nutzlast": "category und summary als Sucheingabe für die Knowledge Base" }, { "von": 2, "nach": 5, "nutzlast": "team, durchgereicht bis zum letzten Schritt für assignedTeam" }, { "von": 3, "nach": 4, "nutzlast": "Bis zu 10 Dokumenten-Chunks (Text, Dateiname, URL, Relevanz-Ranking)" }, { "von": 4, "nach": 5, "nutzlast": "suggestions, confidence und related_tickets" } ], "voraussetzungen": [ "Ticket-System mit API-Zugang und Schreibrechten für Kommentare und Zuweisungen (Zendesk, Freshdesk, Jira Service Management oder vergleichbar)", "Knowledge Base mit gepflegten Lösungsartikeln, durchsuchbar über die jeweilige API (Notion, Confluence, GitBook oder vergleichbar)", "Feste Zuordnung von Kategorie zu zuständigem Team (Hardware, Software, Netzwerk, Zugang)", "LLM-Zugang per API-Key (OpenAI, Anthropic oder Google)" ], "fehlerbehandlung": [ { "fall": "Fehler in einem beliebigen Schritt (Trigger, Kategorisierung, Knowledge-Base-Suche, Lösungsvorschlag oder Ticket-Update)", "massnahme": "Jeder Knoten ist im Originalexport auf die Fehlerstrategie stop gesetzt: Der Lauf bricht komplett ab, es gibt keinen automatischen Retry und keinen Fallback-Pfad. Ergänzen Sie eigene Wiederholungs- oder Eskalationslogik, wenn ein vollständiger Abbruch bei einzelnen Tickets nicht tragbar ist." }, { "fall": "Kategorisierung insgesamt ungenau", "massnahme": "Den Prompt des Kategorisierungs-Agenten um Beispiele je Kategorie erweitern." }, { "fall": "Keine Treffer in der Knowledge Base", "massnahme": "Prüfen, ob die Suche zu eng gefasst ist oder die Indexierung des ausgewählten Knowledge Folders unvollständig ist. Der Lösungsvorschlag-Agent liefert in diesem Fall trotzdem allgemeine Best Practices für die erkannte Kategorie." }, { "fall": "Falsche Team-Zuweisung", "massnahme": "Die Team-Namen und Zuständigkeiten im Prompt des Kategorisierungs-Agenten validieren. Die Teamzuordnung entsteht dort als Freitextfeld team, es gibt keine separate Mapping-Tabelle im Ablauf." }, { "fall": "API-Authentifizierung fehlerhaft", "massnahme": "Tokens und Schreibrechte für Ticket-System und Knowledge Base prüfen." }, { "fall": "Kommentar und Zuweisung erscheinen nicht im Ticket-System", "massnahme": "Der letzte Schritt bereitet category, assignedTeam, comment, confidence und ticketId nur strukturiert auf. Die tatsächliche Schreibaktion ins Ticket-System, also Kommentar speichern und Team zuweisen, ist im Originalexport nicht enthalten und muss als zusätzliche Aktion mit Schreibrechten ergänzt werden." } ], "platzhalter": [ { "name": "category", "quelle": "Ausgabefeld des Kategorisierungs-Agenten: Hardware, Software, Netzwerk oder Zugang" }, { "name": "team", "quelle": "Ausgabefeld des Kategorisierungs-Agenten: Name des zuständigen Teams, muss im Prompt hinterlegt sein" }, { "name": "summary", "quelle": "Ausgabefeld des Kategorisierungs-Agenten: kurze Zusammenfassung des Problems" }, { "name": "maxResults", "quelle": "Konfigurationsfeld der Knowledge-Base-Suche: Anzahl der Treffer, Bereich 1 bis 50" }, { "name": "suggestions", "quelle": "Ausgabefeld des Lösungsvorschlag-Agenten: formatierte Liste der Lösungsschritte" }, { "name": "confidence", "quelle": "Ausgabefeld des Lösungsvorschlag-Agenten: hoch, mittel oder niedrig" }, { "name": "related_tickets", "quelle": "Ausgabefeld des Lösungsvorschlag-Agenten: Anzahl gefundener ähnlicher Fälle" }, { "name": "assignedTeam", "quelle": "Ausgabefeld des abschließenden Code-Schritts, übernommen aus team" }, { "name": "comment", "quelle": "Ausgabefeld des abschließenden Code-Schritts: formatierter interner Kommentartext" }, { "name": "ticketId", "quelle": "Ausgabefeld des abschließenden Code-Schritts, durchgereicht aus dem Webhook-Trigger" } ], "herausgeber": "Voyage Digital", "version": "2.0", "stand": "2026-07-26" }, "itemIcon": "ticket", "recommended": [ { "slug": "antwort-schleife", "category": "workflow", "name": "The Reply Loop", "teaser": "Searches the FAQ knowledge base for every new support ticket, drafts a personalized reply, and hands it to a human for review before it goes out.", "domaene": "Kundensupport & Service", "itemIcon": "reply" }, { "slug": "selfservice-staffel", "category": "workflow", "name": "The Self-Service Relay", "teaser": "Automatically analyzes incoming customer requests, searches the FAQ database for matching articles, and decides based on the solution probability: a self-service reply to the customer or a support ticket with full context to the team.", "domaene": "Kundensupport & Service", "itemIcon": "life-buoy" }, { "slug": "feedback-route", "category": "workflow", "name": "The Feedback Route", "teaser": "A daily automated report that sorts incoming customer feedback by sentiment and topic and immediately flags critical cases to the team.", "domaene": "Kundensupport & Service", "itemIcon": "message-square-quote" }, { "slug": "ticket-route", "category": "workflow", "name": "The Ticket Router", "teaser": "A workflow that automatically sorts incoming support tickets by category and urgency, assigns them to the right team, and prepares a personalized first reply.", "domaene": "Kundensupport & Service", "itemIcon": "route" }, { "slug": "wissens-kette", "category": "workflow", "name": "The Knowledge Chain", "teaser": "A weekly workflow that clusters the most frequent topics from closed support tickets into reviewed FAQ drafts, so the knowledge base stays current without manual research.", "domaene": "Recherche & Wissensmanagement", "itemIcon": "book-open-text" }, { "slug": "cognigy-ai", "category": "tool", "name": "Cognigy AI", "teaser": "Düsseldorf based enterprise platform for conversational AI, with multilingual voice and chat agents for customer service in regulated industries.", "domaene": "Kundensupport & Service", "logoFile": "cognigy-ai.svg", "logoEinzug": 1 }, { "slug": "octonomy", "category": "tool", "name": "Octonomy", "teaser": "Octonomy is a platform for autonomous support agents that handles customer inquiries in e-commerce and SaaS independently and escalates complex cases to a human team.", "domaene": "Kundensupport & Service", "logoFile": "octonomy.png", "logoEinzug": 0.78 }, { "slug": "parloa", "category": "tool", "name": "Parloa", "teaser": "Parloa is a German voice AI platform that gives enterprise customer service teams in regulated industries phone and chat agents with deep CRM and ERP integration.", "domaene": "Kundensupport & Service", "logoFile": "parloa.png", "logoEinzug": 0.78 }, { "slug": "faq-kurator", "category": "assistent", "name": "The FAQ Curator", "teaser": "Answers recurring HR, IT, and customer service questions from your stored FAQ documents, cites the source for every answer, and escalates cleanly to a human whenever the answer is not covered there.", "domaene": "Kundensupport & Service", "itemIcon": "help-circle" }, { "slug": "einwand-drehbuch", "category": "skill", "name": "The Objection Playbook", "teaser": "Classifies a customer objection psychologically and delivers three ready-to-use response strategies with concrete phrasing instead of a generic stock answer.", "domaene": "Vertrieb & Outreach", "itemIcon": "shield-question" } ] }