An iMessage help line for event attendees, answered by AI

Who uses an iMessage help line at events?

An iMessage help line suits conference organizers, trade-show producers, and meetup and workshop hosts whose attendees ask the same things all day: where a session is, the Wi-Fi password, when the next shuttle leaves. Not every attendee installs the event app, and the queue at the help desk grows at every break.

ZeroSettle is an iMessage API: a real phone number on real iMessage, running on Apple hardware we operate for you, that your code texts customers from and hears back on through a bearer token and signed webhooks. For a help line, the hosted AI agent replaces that code, as on a hotel's guest line or an AI receptionist; a restaurant waitlist starts from a QR code too. More are on the use cases hub.

How an iMessage event help line works, step by step

Attendees scan the code on their badge and text the event's number, and ZeroSettle's hosted agent answers from your business context while staff watch.

1. Set the line's name and photo to the event's, such as “BuildConf 2026”. ZeroSettle shares that card with the line's first reply.
2. Write the business_context (agenda, room map in words, Wi-Fi, shuttle times, lost and found) and switch the line group to Managed by AI Agent.
3. Print the number and an sms: QR code on badges and signs: “Text us anything”. Every conversation starts with the attendee's text.
4. The agent answers everything they sent in one go, as up to three texts with the typing bubble first, at one AI credit per text.
5. Staff watch every thread live in the console but can't send from an agent-managed line (403 managed_by_zerosettle), so the context routes anything unlisted to the registration desk.

Conversations are one-to-one: ZeroSettle has no group chats, so nobody sees another attendee's number and nobody can reply-all.

Plan for room changes. Because an agent-managed line refuses your sends, a room change goes into the business context, not out as a message. To text day-of updates to everyone who wrote in, run the line as Managed by API with your own model on the webhook, so your code can send them as well as answer (timing under limits below).

Example conversation: an attendee texts the help line

BuildConf 2026 is fictional; its agent answers from the context shown under setup.

Setup: a badge QR code plus the hosted agent

Setup is a QR code, a business context and one API call (or the same settings in the console).

The badge QR code

Encode sms:+18055550100 (your number, with country code) as a QR code and print the digits beside it. Apple's sms: format takes the number alone and launches Messages, and the iPhone Camera app shows a scanned QR code as a link to tap. Skip the prefilled body the support docs show, so the first text is the attendee's real question. On Android the code opens the SMS app, so print the help desk's location too. More in getting customers to text first.

The business context

The business context is where the agent gets every fact about your event, and it never invents one the context doesn't give: up to 40,000 characters (about 14,000 tokens), used as written. Anyone with a badge, or a photo of one, can text the number, so include nothing you wouldn't print on a sign.

You are BuildConf 2026's automated help line. Say you are the automated
assistant in your first reply. Answer in one short text when you can.
Times are Pacific.

CHANGES TODAY: 2:00 PM "Agents in production" panel moved from Room 204
to Hall C.
THURSDAY: 10:00 AM AI keynote, Hall B (second floor, escalators past
registration). 11:30 AM breakouts, Rooms 201-210 (second floor, east).
LOST AND FOUND: registration desk, level 1, until 6 PM.
WI-FI: BuildConf-Guest, password build2026.
SHUTTLES: downtown hotels every 20 min, 7-10 AM and 4-8 PM, from Pickup A
(north entrance). Say "pickup", never "stop".
Anything else: the registration desk. Emergencies: call 911.

If the agenda doesn't fit, drop abstracts first. A save over the limit is refused with 422 invalid_request and the old context stays, so the agent keeps giving the old room: check every save.

Switch the line group to the agent

This call hands the line group to the agent and turns off two sales defaults: a goal that invites people to buy, and the default trigger, a next-morning follow-up to anyone who goes quiet, which at an event would text nearly every attendee who asked anything. It also sets timezone to the event's zone, because the agent is told the current time in that zone (America/New_York by default) and “starts in 12 minutes” depends on it.

# agent_on.py: run with  python3 agent_on.py  (again after every context edit)
import os, requests

BASE = "https://portal.zerosettle.io"
r = requests.put(f"{BASE}/api/imessage/agent",
                 headers={"Authorization": f"Bearer {os.environ['ZEROSETTLE_TOKEN']}"},
                 json={"managed_by": "zerosettle",
                       "business_context": open("buildconf-context.txt", encoding="utf-8").read(),
                       "goal_enabled": False,
                       "triggers": [],                  # no follow-ups
                       "timezone": "America/Los_Angeles"},
                 timeout=30)
if not r.ok:
    raise SystemExit(r.json()["error"]["message"])   # e.g. a context over the limit
print(r.json().get("ai_credits"))

The console's Messaging tab has the same settings, documented in the AI agent docs.

Limits and consent for an event help line

On the day, an event help line is limited by AI credits and by the line sending one text at a time; ZeroSettle's new-contact limit doesn't apply because attendees text first.

Credit math. 400 attendees with an exchange like Sam's (three agent texts) use 1,200 credits, more than Pro's 1,000 a month, so buy a pack before doors open. At zero the agent goes quiet: texts still land in the inbox, but the agent doesn't go back to answer them. Answer yourself in the console's low-credit banner hands the line back to Managed by API, where staff can answer from the console without using credits.

Break math. The agent's texts wait their turn on the line like any other send, at normally 2–5 seconds each. If 40 attendees text as a session lets out and every answer is three texts, the last answer goes out about 4–10 minutes later; with one-text answers, about 80–200 seconds. That is why the context asks for one short text.

Update math. Under Managed by API, day-of updates also go out one at a time: 150 attendees take about 5–12 minutes. Send them one after another, because a send that waits about 12 seconds behind others returns 429 line_busy (retry after 2 seconds). On every plan, a second update to an attendee who hasn't replied since your answer is refused with 403 awaiting_reply.

Consent. Attendees initiate, so answering them is expected; if you plan day-of updates, say so on the sign and keep them to logistics. Post-event marketing needs a separate opt-in: in the US, marketing texts need written consent (47 CFR 64.1200 defines “prior express written consent”). ZeroSettle's terms require disclosing AI unless contextually evident, hence the agent's opening words. Consent rules vary by place and purpose; this isn't legal advice.

When ZeroSettle is not a fit for an event help line

ZeroSettle is the wrong tool when an event must reach everyone at once or many attendees lack iMessage.

• You need an announcement channel. There are no group chats, and an agent-managed line can't send your announcements.
• You need instant updates to thousands. At 2–5 seconds a message on one line, 2,000 attendees take roughly one to three hours; use email or the event app's push notifications.
• Many attendees carry Android phones. iMessage reaches only iPhone and other Apple users (Apple Support), and ZeroSettle has no SMS fallback.
• You want AI answers and staff replies in one thread, as a code-of-conduct or safety report needs. Staff can read an agent-managed line but not send from it, so give reports a staffed channel and name it in the context.

How much does an iMessage event help line cost?

An event help line costs the ZeroSettle plan plus any AI credit packs a busy event needs.

For the 400-attendee event above, Pro plus one 1,000-credit pack is $99 and covers 2,000 agent texts, if the month's included credits are untouched. Packs (1,000 credits for $50 or 5,000 for $200) never expire, so leftovers stay on the account for the next event; the month's included credits don't carry over. Growth's extra numbers share the load only if you print different numbers, for example one per badge type, because each attendee is answered from the number they texted.

Pro is self-serve, with no setup fee, unlimited messages with people who message you first and a 30-day money-back guarantee on the subscription; Growth starts with Book a demo. See ZeroSettle pricing, the plans and the API docs, or get a number.