Case Study - Self-service AI demos, claimed in one click

Overview

As part of launching amazee.ai, a private, self-hosted AI infrastructure offering, amazee.io wanted prospects to try AI-driven editorial workflows, intelligent search and categorisation in Drupal for themselves. Editorial AI demonstrates poorly on a slide, and every manually built demo is an engineer not building the product.

Demos had to deploy in the US, Europe or Australia, arrive pre-seeded with LLM endpoints and vector database integrations, send their own lifecycle emails, feed the CRM, and clean themselves up afterwards.

What we did

We co-developed Polydock, an open-source orchestration engine that automates demo environments at scale, and extended it with a concept we called the hot demo.

Pre-provisioned instances sit idle and ready to be claimed. When a prospect requests a demo in a region where one is waiting, it is assigned instantly. When there is not, Polydock builds a new one from a snapshot image, which is slower but still nothing like a manual setup.

Around that: region-aware deployment on Lagoon-backed Kubernetes, one-time login with credentials delivered by browser and email, automated lifecycle messaging in branded templates, HubSpot integration capturing every trial, and automatic expiry and cleanup after seven days.

The outcome

A prospect launches a Drupal AI demo in one click, in minutes rather than hours, across three continents with data residency options. Sales and engineering handle no manual demo setup. Because every trial lands in the CRM, demo traffic became measurable leads that can be segmented and followed up instead of anonymous interest.