Old-style chatbots work with predefined menus and keywords and fail when a question does not fit the pattern. New AI-powered customer assistants understand natural language, answer from the company's knowledge base and hand the conversation to a person when needed.
What they can handle
- Frequently asked questions: opening hours, prices, delivery times, return conditions,
- Order and shipping status (with system integration),
- Appointment and reservation requests,
- Product and service recommendations,
- First-step troubleshooting,
- Collecting requests and routing them to the right team.
Which channels?
Website chat, WhatsApp Business, Instagram messages and email are the most common. What matters is collecting conversations from every channel in one panel the team can monitor.
Designing a good assistant
- Prepare the knowledge source: Current FAQs, price and policy documents should be the assistant's single source of truth.
- Set boundaries: Which topics will it answer, and which not? Discounts, complaints and legal matters should go to people.
- Easy human handover: Customers can reach a person at any time without losing the conversation history.
- Brand voice: Define tone, length and form of address.
- Personal data: Inform customers, avoid asking for unnecessary personal data and set retention periods for conversations.
Metrics to track
- Share of conversations resolved without human handover,
- First response time and total resolution time,
- Customer satisfaction score,
- Questions the assistant could not answer (the most valuable input for improving the knowledge base).
How to start
Use the assistant first for information and request collection only. When results and feedback are positive, extend it to tasks that need system integration, such as order lookups and appointments.
ujans builds custom customer assistants with knowledge base preparation, website and messaging integrations and human handover processes.