IBM Watson Assistant

You are a customer-support copywriter for an enterprise chatbot. Produce three concise escalation handoff messages the virtual agent will send when transferring a conversation to a live agent. Constraints: 1) Provide three distinct tones (empathetic, formal, technical); 2) Each message must be ≤40 words and include a single-line summary of the customer's issue; 3) Include a one-sentence brief the bot should attach for the live agent. Output format: JSON array of 3 objects with keys {tone, bot_message, agent_brief}. Example object: {"tone":"empathetic","bot_message":"...","agent_brief":"..."}.

You are an NLU content specialist. Generate 12 diverse, realistic user utterances to train the intent "report_lost_card" for a banking chatbot. Constraints: 1) Each utterance must be unique and under 12 words; 2) Include slot variants for card_type (debit/credit) and last_four digits (use X placeholders, e.g., 1234); 3) Add 3 negative examples (similar-sounding but different intent). Output format: JSON with keys {"positive_examples":[], "negative_examples":[]}.

You are an assistant that outputs production-ready dialog node templates for Watson Assistant. Create a JSON template for a node named "verify_identity". Constraints: 1) Include keys: id, name, conditions, intents (list), entities (list), context_updates (object), responses (array of objects with type:text and optional discovery_query), and next_step (object); 2) Provide two response variants: primary and fallback; 3) Keep placeholders where dynamic values are required. Output format: a single valid JSON object. Example responses element: {"type":"text","text":"Please confirm the last four digits of your SSN."}.

You are a knowledge-engineering assistant designing retrieval-augmented responses for Watson Assistant integrated with IBM Discovery. Produce a response template that: 1) Returns an answer paragraph followed by up to two cited excerpts (source name, short snippet, link); 2) Includes a confidence score bracket (High/Medium/Low) and an automatic follow-up suggestion when confidence is Medium/Low; 3) Always includes a fallback CTA when no documents match. Output format: JSON object with keys {answer, citations:[{source, snippet, url}], confidence, follow_up_suggestion, fallback_cta}. Provide one filled example and leave placeholders where query-specific values belong.

You are a senior product manager designing multi-turn conversational journeys for a customer-support assistant. Produce three complete flows (billing dispute, password reset, product setup) in YAML-like structured format. For each flow include: 1) entry conditions (intents/entities), 2) required data slots with validation rules, 3) prompts/messages for each turn, 4) escalation criteria (when to handoff), 5) example user-bot utterance pairs (2 per flow). Follow this few-shot example style: ExampleFlow: {entry_intent: refund, slots:[order_id|required|regex], turns:[{bot:..., user_example:...}], escalation: 'after 3 failed validations'}. Output exactly three flows in the same structure.

You are a cloud migration lead preparing a migration plan for moving Watson Assistant from IBM Cloud public to IBM Cloud VPC or on-prem deployment. Produce a step-by-step migration plan containing: 1) phases (assessment, staging, migrate, validate, cutover, rollback), 2) deliverables per phase, 3) timeline estimates (in days) and resource roles, 4) key risks with mitigations, 5) test cases (functional, performance, security, compliance), and 6) a rollback decision tree. Output format: numbered phases with nested bullet lists; include a short executive summary (3 sentences) at top.

Compare IBM Watson Assistant with Google Dialogflow, Microsoft Bot Framework, Amazon Lex. Choose based on workflow fit, pricing, integrations, output quality and governance needs.