Home›Sunday Sessions›Before You Text
Sunday Sessions · Specialist Edition
What This Is
iMessage. SMS. WhatsApp. Telegram. These are the most personal communication channels a person has. When your AI agent lives there, the standard for trust is not the same as a chatbot on a website. It is higher. The user did not seek out an interface. The message arrived in the same place as their family, their doctor, their oldest friend.
What happens when your agent is challenged? When a user asks directly: "Am I talking to a human?" What does it say? What does it do when a support ticket goes unresolved for two weeks while the AI continues to tell the user everything is fine? What happens when it produces a name — a human name — under pressure?
These are not edge cases. They are the predictable failure points of every text-based AI agent that was built without a trust framework. This session is how you build one before your users find it for you.
This is not theory. The framework you will build in this session comes from direct, documented experience — 311 automations built and deployed on a text-based platform, a complete breakdown of the support infrastructure, and a firsthand account of exactly what happens when the trust layer is missing. Read the full account here.
Identity disclosure — designed for your agent, your channel, your users
Every text-based AI agent needs to know exactly how to introduce itself, what to say when challenged, and what it is never permitted to claim. "hi there, this is human support" is not a gray area — it is a false claim that travels at the speed of a screenshot. You will leave this session with the exact language your agent uses to identify itself, and the guardrail that ensures it never crosses the line regardless of how the conversation is steered.
A working escalation path — not a ticket queue that goes untouched
A support ticket that sits unassigned for 17 days while the AI tells the user everything is under review is not a support system. It is the appearance of one. You will design a real escalation path — who gets notified, when, by what trigger, and what "unresolved after X hours" actually means in your system. Infrastructure, not headcount.
Human-in-the-loop protocol for the decisions that matter
Financial notifications. Creator payouts. Anything touching a user's livelihood, their trust, or their data. These are not AI decisions. You will define exactly where the human stays in the chain — not as a feature, but as the architecture — so that when something goes wrong, a real person is already assigned, already watching, and already empowered to act.
Red team checklist — find the failure before your users do
Ask your agent directly: "Are you human?" Ask it to give you a name. Push it past the edge of its training. If it can be steered into claiming a human identity, your users will find that boundary — not in a test environment, but when something has already gone wrong and they are looking for answers. You will leave with a testing protocol that surfaces the failure before launch, not after.
Who This Is For
You do not need to be a developer. You need to be the person responsible for what that agent says — and what happens when it says the wrong thing.
You are building on iMessage, SMS, or WhatsApp
You have a working agent or you are close. You have not yet defined what it says when challenged, what it escalates, or what it never does. This session closes that gap before launch.
You are advising a team building one
You are a consultant, a product lead, or an advisor on a project that involves a text-based AI agent. You need a framework you can hand to the team and defend to the stakeholders.
You already launched — and something went wrong
Your agent said something it shouldn't have. A user was confused about whether they were talking to a human. A ticket went unanswered. You need to retrofit the trust layer correctly, not patch it.
Why This Credential Matters
Apple's approval for any agent on Messages for Business requires two things: verified live support capability and clear AI identification. Both are exactly what most teams get wrong.
Getting approved for iMessage — or any major messaging platform — is not just a technical milestone. It is a trust certification. The platforms require it because users demand it. An agent that fails on identity disclosure or support responsiveness does not just lose a user. It loses the platform relationship. The trust framework you build in this session is the same one that satisfies Apple's requirements, protects your users, and keeps your business on the right side of the line.
What This Is Not
A general AI overview
This session is not about AI broadly. It is about text-based AI agents specifically — the trust layer, the identity architecture, and the escalation design. Nothing generic.
A build session
This session does not produce code. It produces the decisions that make the code trustworthy. What to build follows from what you decide here.
A recording
This session is not recorded. What is decided in the room stays in the room — which is exactly why people bring their real problems.
A sales pitch
Nothing in this session is designed to sell you a service. If you leave and want something built, that conversation happens separately on your terms.
Refundable
All sales are final. The seat is yours the moment you reserve it. Plan to be there and ready to work.
Frequently Asked Questions
What is the Before You Text session?
Before You Text is a focused 90-minute working session for founders and small business owners who are building or deploying a text-based AI agent — on iMessage, SMS, WhatsApp, or Telegram. The session covers identity disclosure design, escalation path architecture, human-in-the-loop protocol, and red team testing before launch. You leave with a working trust framework specific to your system.
Who is this session for?
This session is for anyone building or deploying an AI agent that communicates with users over text. Founders, small business owners, product teams, consultants. If your agent lives in iMessage, SMS, WhatsApp, or Telegram and you have not yet defined what it says, what it escalates, and what it never does — this session is for you.
What do I walk away with?
Four things: a defined identity disclosure protocol for your agent, a working escalation path design, a human-in-the-loop framework specific to your system, and a red team checklist to test your agent before it reaches your users.
Why does the trust layer matter so much?
An AI agent that claims to be human — even once — is a liability, a trust failure, and in regulated industries, a legal exposure. Apple's approval for any agent on Messages for Business requires verified live support and clear AI identification. These are not optional. Read the firsthand account of what happens when these fail.
How much does it cost?
Before You Text is $597 per seat. All sales are final. Sessions run 90 minutes live on Google Meet. Seats are limited and filled first come, first serve.
The Details
Before You Text runs on a rotating schedule. Dates are announced after you reserve — the way in is to secure your seat and you will be notified of the next available session. Seats are limited. First come, first serve. No exceptions.
Format
Live · Google Meet
Duration
90 minutes · Focused working session
Seats
Limited · First come, first serve
Camera / Mic
Optional · Zero pressure
Recording
Not recorded · Live only
Schedule
Rotating · Notified upon registration
Deliverable
Trust framework · Yours to keep
Refund Policy
Non-refundable · All sales final
$597
per seat · 90 minutes · trust framework included
Reserve Your SeatNon-refundable · All sales final · One session per seat
Hello. I'm Soli — I know this business inside and out. What can I help you figure out today?