Will customers know it's AI? Trust, disclosure and keeping a human in the loop

Short answer: customers mostly care about two things: a correct answer and a way to reach a person. Get those right and AI help builds trust instead of costing it.

What customers say

• 69% are comfortable with brands using AI to give faster care (Sprout Social, 2025, vendor data).
• 64% would prefer companies didn't use AI for customer service, and the top worry is that it gets harder to reach a person (Gartner, 2024).
• Nearly 75% want to know when they're talking to an AI agent, and 45% are more likely to use one if there's a clear way to reach a human (Salesforce State of the AI Connected Customer, 2024).
• 50% are put off by generic or templated review replies (BrightLocal, 2026).

These look contradictory until you see the pattern: people don't mind help from AI. They mind dead ends, wrong answers and copy-paste replies.

Four rules that keep trust

• Answer from real facts. Give the AI your prices, hours and policies. A confident wrong answer does more damage than a slow one.
• Always leave a way to reach a person. Hand over complaints, refunds and anything emotional.
• Sound like you. Use your words and your sign-off, and cut the filler.
• Be open when it's a bot. If a reply goes out with nobody checking it, consider saying so. Meta recommends disclosing automated chats, and some laws require it (Meta).

Approve first, automate later

Most teams start with the AI suggesting replies that a person sends, then move to one-click approval, then let it answer simple buying questions by itself. In AeroReply that's the Suggest, Approve and Auto modes, plus a minimum intent score for anything sent automatically (how it works).

What never to hand to AI alone

• Refunds, compensation or legal questions.
• Health, safety or anything regulated in your industry.
• Angry customers. Reply personally, then fix it in private.