Answer missed calls after hours
A missed call gets a text back. In your opening hours the text is plain. Out of hours this n8n workflow opens a short SMS chat that answers only from the FAQ you write in your own Google Sheet, takes the caller's name and what they need, and emails you the lead.
Coming soonNot on sale yet
The question
this build answers
How do I text back missed calls and answer simple questions by SMS when I am closed, without the AI making things up?
Point your Twilio number at this n8n workflow. A missed call gets a text. In hours it is a plain text-back. Out of hours it invites a reply, and OpenAI only picks which rows of your FAQ tab answer the question. The workflow sends those rows word for word and never writes its own answer. When it has a name and a need, or the text is urgent, it adds a row to your Leads tab and emails you.
From the first trigger
to the finished job.
The working version, step by step.
Twilio reports the missed call
Twilio posts each missed call to the workflow, which checks your Hours tab for your weekly hours, holiday dates and overnight spans.
A text goes back
In hours the caller gets a plain text-back. After hours the text invites them to reply.
Answers come from your FAQ only
OpenAI only picks which FAQ rows answer the question, and only for a number that rang you inside the last 24 hours, a default you can change. The workflow sends the answer text of those rows word for word. If no row answers it, or the model fails, the caller gets a fixed message and the chat is handed to you. Only the first 40 FAQ rows are used, and a text is at most 320 characters.
The lead reaches you
With a name and a need, or an urgent text, it adds a row to your Leads tab and emails you. An urgent lead alerts at once. The other overnight leads arrive in one digest when you open.
STOP, repeat callers and odd numbers
It handles STOP, START and HELP. Not every call gets a text. A withheld number, a landline, a number on your ignore list and a caller who has opted out get none. A caller who rings five times in ten minutes gets one text, and there are caps per caller. Once a chat is handed over, more texts from that caller go onto the same lead and get no reply until the idle time you set has passed. Every turn is logged in your Conversations tab.
Not on sale yet
After-Hours Receptionist for Missed Calls (n8n)
n8n is still reviewing this template for its own library. It goes on sale here once its test is on record. The zip holds one workflow file to import into your own n8n and a README with every setting.
Coming soonWhat we have run, and what we have not
Every Code node runs offline as shipped, 341 checks in all. On a real n8n instance with a real Google Sheet, 20 of 20 live cases passed, on the file in the zip and the build just before it. Real OpenAI ran on six prompts, and every reply was FAQ text word for word or a fixed message.
We did not run the Twilio send live, because we had no Twilio account. A webhook of our own stood in for Twilio and recorded each text, so no real text message was sent. The owner alerts went through a stand-in mail service and not through Gmail. Twilio's own delivery and the Gmail alerts are not proven on this file.
What you need
n8n 1.x or 2.x, a Google account for Sheets and Gmail, a Twilio number that can send SMS and an OpenAI API key. Slack is optional.
Twilio charges for each text, and each chat turn is one small OpenAI call.
Check your local rules before you switch to live. In the UK that means PECR, and in the US Twilio requires A2P 10DLC registration before business texts are delivered.
What goes to OpenAI
The caller's words, any name and need they give, and your FAQ rows go to OpenAI with each turn. The caller's phone number does not. The model never writes a word the caller reads.
It starts in preview
In preview nothing is texted to a caller. Each text it would send arrives in your inbox, and every row is still written. The AI still runs in preview, so a chat costs the same as in live mode.
Limits worth knowing
It is not a voice agent. Nothing here answers a call, records a message or reads a voicemail.
The key in each webhook address is a shared secret. There is no Twilio signature check.
It reads the whole Conversations tab for each text, so archive old rows every few months. Hundreds of chats a day need a real database.
A failed text is not resent. You get an alert and the lead stays New.
The model reads English best.
The fit matters.
Who should not build this
If you want calls answered by voice, this does not do it.
If you do not have a Twilio number that can send SMS, it cannot run.
Skip it if callers' messages must not go to OpenAI, because each chat turn sends them there.
Yours to keep
What lands
on your desk.
- One n8n workflow file to import into your own n8n
- A README with the sheet tabs, every setting, the limits, and what was and was not tested
A few useful
answers.
Can I buy it today?
You cannot buy it yet. It goes on sale here once its test is on record.
Can the AI make up an answer?
The model cannot write to the caller. It only picks which FAQ rows answer the question, and the workflow sends those rows word for word, or a fixed message when no row fits.
Does it answer the phone?
It does not answer calls. It texts back after a missed call, and nothing here reads a voicemail.
What do I need to run it?
n8n 1.x or 2.x, a Google account, a Twilio number that can send SMS and an OpenAI API key.
A little further down
the same path.
Related builds for the rest of your business.
Make this work for you
Your tools.
Your working version.
Ask for a build to fit
Quoted to your job
Tell us what your version needs to do. One line is enough to start.
Start a brief