Should an AI answer your customer texts?

Every vendor in this category now has an AI story. Auto-reply, lead qualification, agents with human handoff. The technology works better than it did two years ago, which makes the question harder rather than easier — it is no longer 'can it', it is 'should it'.

Start from what the channel is for

A small business wins on this channel because the customer is talking to a person they have met, from a number they recognise. That is the entire competitive advantage — it is the one thing a national chain cannot buy. Automating it away to save fifteen minutes a day is a bad trade, and it is a trade many businesses make without noticing.

Disclose it

Not because a regulation necessarily requires it in your jurisdiction, though it may. Because customers work it out, and discovering it after the fact feels like being tricked by a business they trusted. 'This is our automated assistant — a human will pick this up shortly' costs you nothing and protects the relationship.

Handoff is the whole design

The failure mode is not a wrong answer, it is a customer trapped in a loop with no way to reach a person. Every automated thread needs an escape hatch that works on the first attempt.

const ESCALATE = /\b(human|person|agent|manager|speak to someone|real person)\b/i;
 
export async function handleAgentTurn(thread: Thread, text: string) {
  // Explicit ask always wins, immediately.
  if (ESCALATE.test(text)) return escalate(thread, "customer_requested");
 
  // Confidence floor: do not guess at a customer.
  const reply = await agent.respond(thread, text);
  if (reply.confidence < 0.7) return escalate(thread, "low_confidence");
 
  // Anything emotional goes to a person, regardless of confidence.
  if (reply.sentiment === "negative") return escalate(thread, "negative_sentiment");
 
  // Hard cap: no thread runs unsupervised forever.
  if (thread.agentTurns >= 4) return escalate(thread, "turn_limit");
 
  return send(thread, reply.text);
}

The rule of thumb

Use AI to make sure nobody waits, not to make sure nobody is answered. Acknowledge instantly, gather context, draft the reply — then let a person send it. That combination gets you the response times of automation and the trust of a small business, which is the whole point.

And whatever the agent sends, do not let it compose its own URLs. Give it a tagging helper, or your attribution will drift within a week — see UTM tagging.