An automated text response is a reply your system sends when someone texts you first, so they get an answer in seconds even when nobody is free. Set up well, it stops new enquiries going cold overnight and at weekends, answers the simple questions instantly and hands the rest to a person with context. This article answers how to build automatic replies to inbound texts on the levelpartner.ai platform, from a simple acknowledgement to an AI agent that holds a short conversation.
Three levels of automated reply
Pick the level that matches your volume and risk. Most firms start at the first and move up.
- Acknowledgement. A fixed reply that confirms receipt and says when a person will respond.
- Keyword replies. Different answers for common words: BOOK, PRICE, HOURS, STOP.
- AI agent replies. An AI agent reads the message, answers from approved information, offers booking times and hands over when it reaches its limits.
The first is safe and takes an hour. The third is the most useful and needs the most care.
Before you start
- Inbound texts arriving in Conversations from your business number.
- Agreed opening hours, so every acknowledgement can say when a person will reply.
- Owners assigned to contacts, so a person picks up every conversation.
- For AI replies, a short document of approved answers: services, areas covered, how booking works.
- A clear list of topics the automated reply must never discuss, such as pricing advice or complaints.
How to set it up
- Go to Automations, then Workflows, and create a new workflow for inbound texts.
- Click "+ Add New Trigger" and select "Customer Replied".
- Enter a Trigger Name such as "Inbound text: BOOK keyword".
- Click "+ Add Filters" and set Reply Channel to SMS.
- Add a Contains Phrase filter for BOOK, so only those texts enter this workflow.
- Click "Save Trigger".
- Add a "Send SMS" action that replies with the owner's booking link.
- Name the action, write the reply with {{contact.first_name}}, and send a test SMS to your own phone.
- Save the workflow, test it from a personal phone, then publish it.
Writing the acknowledgement
The acknowledgement is the simplest message and the one most often written badly. "Thank you for your message. We will be in touch" gives the customer nothing. Name a time and an option.
"Thanks, Jo. The team is back at 8:30 tomorrow and Alex will reply first thing. If it's about a viewing, you can book directly here: [link]."
That message sets an expectation and lets a ready buyer act now. For more on short-form wording, see automated message text.
Suppose the office gets about 60 inbound texts each weekend, mostly about viewings, and Monday morning goes on working through them. If an AI agent replied and booked viewings for, say, half of those conversations, around 30 would be settled before Monday. The other 30 would arrive with a summary in the thread. Staff would pick them up with context instead of starting from scratch. The volumes are assumptions.
Where AI replies help and where they do not
An AI agent is good at availability questions, simple service questions and booking. It is not suitable for advice. A mortgage broker's agent must not comment on what a customer can borrow. An agency's agent must not quote a project price. Write those limits into its instructions and test them deliberately by asking forbidden questions yourself.
The agent should always end by telling the customer what happens next and who will follow up. A conversation that stops without a handover feels like a dead end. The instant follow-up patterns in automated marketing show how replies connect to wider sequences, and the missed call text back example covers the reply for calls nobody could pick up.
What it replaces and the weekly time
Automated responses replace the out-of-office habit of checking the business phone at night, and the Monday backlog. Weekly effort is a review of a sample of automated conversations, perhaps 30 minutes, to catch any answer the agent got wrong and adjust its instructions.
The mistake most people make
Letting the automated reply pretend to be human when asked. If a customer asks whether they are talking to a person, the honest answer keeps trust. Configure the agent to say it is an assistant and that a named person will follow up.
When not to auto-reply
Do not auto-reply to existing clients in the middle of sensitive work, such as a sale falling through or a complaint, where a fixed message can sound dismissive. Use a tag to exclude those contacts. Do not auto-reply during business hours if your team can answer within minutes; a person is better.
Troubleshooting
Contacts get two replies to one message. Two workflows both start on Customer Replied with overlapping filters. Combine them or tighten the filters so only one runs.
The keyword reply never fires. Customers type the word in a sentence. Use the Contains Phrase filter rather than Exact Match Phrase.
The AI agent answers off-topic questions. Its instructions are too open. Restrict it to your approved answers and add an explicit handover rule.
STOP replies get an acknowledgement. Tighten the trigger filters so opt-out words never match, and let the system handle unsubscribes on its own. See automated text for business for consent handling.
Frequently asked questions
Can an automated text response book appointments?
Yes. A keyword reply can send a booking link, and an AI agent can offer specific slots from your calendar and book them in the conversation.
Will customers know the reply is automated?
They should. Sign acknowledgements from the team, and set the AI agent to be honest when asked.
Does the system reply to every inbound text?
Only those that match your workflow conditions. You choose whether it runs out of hours, on keywords, or always.
Can the same setup handle WhatsApp?
Yes. Build a similar workflow for WhatsApp replies, keeping the wording suited to that channel.