AI agents and traditional chatbots
Traditional chatbots follow fixed scripts or decision trees. AI agents read a shopper's message in their own words and answer from your knowledge. Some can also take actions in your store, such as processing a return.
What to watch for
- Made-up answers, if the AI isn't limited to your own sources.
- Order details shared with the wrong person, if nothing verifies the shopper.
- Costly actions taken with no check, such as an incorrect refund.
- Frustrated shoppers, if there's no easy way to reach a person.
How to roll it out safely
- Give it your policies, product data and common answers.
- Test it on real questions before shoppers see it.
- Start with questions and order status.
- Add order actions one at a time, requiring approval first.
- Review handed-off conversations, and fill the gaps in your knowledge.
How Counter approaches it
Counter answers from your own sources and live store data, verifies shoppers by email before sharing order details, and lets you decide topic by topic whether it acts, waits for approval or hands off. When it isn't sure, it passes the chat to your team.
FAQ
Common questions.
Can AI replace a store's support team?
- It can take on repetitive questions, but your team is still needed for exceptions, judgment calls and upset customers. Plan for a clear handoff.
Is AI customer service safe for refunds?
- It can be, with controls: verify the shopper, limit actions to your published rules, and require a teammate's approval until you trust the results.