You can open most of this without talking to me.
Anyone can describe what they would build. Below is what I already did, each piece labelled with what it actually is: running with paying customers, open to the public, or a demo you can try. The labels matter more than the screenshots — a demo sold as production falls apart the moment you ask a detailed question, and takes the rest of the story with it.
One of them is on this page: the assistant in the corner is the same system I install for clients, running on my own site and taking my own leads.
A quoting engine that closed the phone call
Every order started the same way: a message asking "how much for 500 badges?". Someone had to be awake, look up the price table, do the maths for that volume and write back. Quotes went out hours later, and the ones that arrived at night went out the next morning — to a customer who had already asked someone else.
- A live quote engine: the buyer picks badge type, quantity and delivery option, and the price moves as they choose — no form, no waiting.
- A volume curve drawn from the shop's real price table, so the discount for ordering more is visible instead of negotiated.
- A written quote generated on the spot, with the terms attached, so nothing has to be re-typed later.
- A design preview where the customer types their own name and organisation and watches the badge change.
The business quotes around the clock without anyone being there, and the conversation that used to start with "how much?" now starts with "I already saw the price, let's order."
A document assistant that has to cite its source
People in debt ask the same twenty questions, and every wrong answer is expensive. A generic chatbot answering from memory is worse than no chatbot: it sounds confident and it can be wrong.
- The platform's own documents turned into an assistant people can ask in plain language.
- Every answer comes back pointing at the document it came from, so nobody has to trust it blindly.
- Hard guardrails on scope: it explains what the documents say and it does not give legal advice — the boundary is enforced in the system, not left to the model's judgement.
A self-service platform where the assistant does the reading and the person keeps the decision. It is open: you can go and ask it something right now.
The front desk that does not close
The patient who decides to book at nine at night finds nobody, and books somewhere else. The front desk spends its day on the phone rescheduling instead of looking after whoever is standing in front of it.
- An assistant that takes bookings over chat and WhatsApp at any hour, in the language the patient writes in.
- Confirmations and reminders that go out on their own, because the appointment nobody confirms is the appointment nobody attends.
- A handoff to a human the moment the conversation stops being about scheduling.
It is a working demo, open without signing up — no clinic is running on it yet, and saying otherwise would be the fastest way to lose the first one.
Everything else, labelled honestly.
Some of these have customers. Most do not yet. The list is useful because it is exact.
About the screenshots. The client systems above serve Spanish-speaking markets, so their interfaces are in Spanish and their prices are in local currency. The images on this page show those same live products with the interface translated and the amounts converted to US dollars at the published rate of the day — the layout, the data and the behaviour are the real thing, untouched. Open the links and you will see them exactly as their customers do.
Ask it about any of these.
The assistant on this page knows every project above. Ask it how one of them was built, or describe the job your team keeps doing by hand and let it tell you whether it is worth automating.