The demo worked. Getting the first school live is a different kind of work.

On Sep 16 I showed the school booking platform again.

This time I demonstrated an AI completing a booking through a phone call. I also walked through the feedback from the previous review and the changes I made.

The core result was encouraging: the product is close enough that we can start putting the first school on production and setting them up, even while Twilio SMS approval is still in progress.

That is a real step. It is also where a different kind of work shows up.

The demo was not the finish line

The earlier review feedback was addressed. The new demo brought a smaller, more specific set of refinements. One of them changed how I think about the voice experience.

The AI was asking for too much information at once.

That is a subtle problem, but it matters. A booking conversation should feel like a person guiding someone through a simple decision, not like a form being read over the phone.

I am narrowing the first part of the call to the essentials: name, preferred date, and preferred time. Rather than listing every upcoming availability immediately, the agent will ask for a date and time preference first, then read the relevant time slots in sequence.

I also decided to drop email capture from the phone call. Voice models are not reliable at catching email addresses, even when someone spells them out letter by letter. That stretch makes the call longer and more frustrating for almost no gain. The school wants people to show up. So the call should stay simple: get them booked. Email can wait. Once SMS is approved, the text becomes the verification check, and that is where we ask for email.

The feedback was not asking for a bigger system. It was asking for a calmer conversation and a cleaner path to “they are coming.”

The phone-call demo mattered because it showed the full experience working together, not just screens or isolated features. Now the work is making that experience simpler for the person on the other end of the call.

Production has its own checklist

Shipping to production and starting school setup does not have to wait on SMS.

What does wait is text messaging. To send texts through Twilio, I need approval for the messaging setup. That process is more than adding an API key.

I need a simple public site for Spray Pixl (the business behind the platform), including basics like a privacy policy. I also need a separate document that explains how people consent to receive texts, plus screenshots from the staging booking widget showing that opt-in flow.

None of this is the fun part of a demo. It is still part of building a real product.

If the platform is going to text people, consent has to be clear, the business needs a credible public presence, and the messaging setup has to meet Twilio’s requirements. Until that lands, production can still host the booking flow. SMS verification and the follow-up email ask wait on approval.

What I am learning

Getting close to production changes the work.

Earlier, the important questions were about workflows. Can the platform support the school’s process? Does the AI booking flow make sense? What does the product still miss?

Now the questions also include operational trust. Can messages be sent responsibly? Is consent documented? Does the business infrastructure hold up when the software starts contacting real people?

Those are not side projects. They are part of the product becoming usable.

What happens next

I am implementing the remaining voice feedback (narrower questions, no email on the call) and completing the Twilio approval materials.

In parallel, I can put the first school on production and start setup so we are not blocked on paperwork for everything.

The immediate goal is straightforward: a production booking flow the school can use, SMS approval so verification and email capture can happen by text, and the operational pieces in place to support it.

Honest status: the first school is ready to start on production. The remaining product feedback is focused. SMS is the main external gate. The work now is turning a strong staging product into something that can operate with real families.