Ask Moto: a production AI riding coach

The problem was never the chatbot

The riding system behind this app has been ten years in the making and has trained more than ten thousand riders. The system is the business. Maker Factory wanted it available to riders between lessons, at the moment a question actually occurs: late on a Saturday night before a Sunday ride, not in a classroom.

Anyone can put a language model behind a chat box. The reason this needed engineering rather than configuration is grounding. A riding coach that invents technique is worse than no coach at all, because a rider will take that advice onto a road at a hundred kilometres an hour. Every answer had to come from that training material, be traceable back to it, and the system had to decline rather than improvise when a question fell outside what the corpus covered.

That constraint drove the architecture. Indexing the material so retrieval returns the right passage for a technique question rather than a vaguely related one. Constructing prompts so the model answers from retrieved context rather than its own training. Handling the cases where retrieval comes back thin, which is where most retrieval augmented systems quietly start making things up.

What shipped

Ask Moto sits inside a complete consumer product rather than standing alone. The app delivers one short coaching lesson a day on cornering, braking and observation, with a streak counter and calendar to build the habit. It carries the streetSKILLS 101 foundation videos, runs monthly challenges with a global leaderboard, and syncs progress across a rider’s devices.

Version 1.0 shipped on 18 May 2026, followed by eight releases through to the end of August covering in-app payment fixes, subscription handling, rider onboarding and discount codes. The app is live and maintained on both stores.

It is a paid product, not a demo

This is the part that matters most, and it is the part I did not control. Ask Moto sits behind a Premium AI Coach subscription, priced separately from the base app subscription. Riders choose to pay for the AI tier specifically.

A great deal of AI work never crosses that line. It demos well, it impresses a steering committee, and it quietly never reaches a customer who values it enough to pay for it. The commercial structure here is third-party evidence that this one did.

Delivery

Design through to release in under two months, solo, including the retrieval pipeline, the mobile application on both platforms, and in-app payments for subscriptions and consumables. Built using agentic development tooling, inside the quality bar a product needs to survive paying users.

Related reading:

• RAG patterns for enterprise AI
• using AI in custom mobile apps
• building mobile apps faster with AI development tools