The Finnish elevator maker KONE developed an AI-driven monitoring and control system that could in real-time analyze any elevator’s status based on over 30 variables, analyze it to predict near and far future wear and tear — and repair things before they break. Powered by IBM Watson IoT the KONE cloud had enough power to handle a million elevators simultaneously, the system was ready to go live but needed a powerful demonstration of 24/7 elevator connections that would appeal to the general public, to media, and target customers. We started out with a simple, threefold aim: Create worldwide buzz for predictive maintenance, make KONE the opinion leader in that sector, and establish KONE as a force to reckon with in AI-based servic

Talking about AI, IoT and Machine Learning would get us nowhere. So we asked a simple question: has anyone ever HEARD machines talking to each other — in real-time? We knew the new KONE service would be able to listen to vast amounts of real-time use data from up to a million elevators — each second. We devised a way to tap into the real-time data feed between KONE elevators and the IBM Watson IoT cloud and created intelligent, human-language discussions from these. Human-language equivalents were created for each elevator and environmental variable and Watson IoT action. Thus creating a language, tone-of-voice, and personality for our elevators & AI — and making it possible for them to have conversations that people could eavesdrop on.

The real-time data supplied by KONE’s sensor packs in 12 elevators was uploaded via a custom API to a real-time human language database that translated the elevator, environmental & AI cloud variables into over 500 status update sentences in human language — spoken live using IBM’s BlueMix voice synthesis. The sentence content, tone and structures varied based on each of the over thirty monitored variables — and the responses by the cloud were similarly given human-equivalent versions. All this, including live audio, was output on our campaign site in real-time following the elevators’ second-by-second status changes. Nobody had ever translated machine-to-machine data into audible, intelligible and shareable conversations. Until now.

In days, Machine Conversations became a pop-culture phenomenon. In only five weeks it reached 327 million listeners and viewers — with e.g. over 350 media articles ranging from USA Today to Forbes to Financial Times to Quartz to BoingBoing. Over 3 million € in earned media in first weeks. The sentiment of coverage was highly positive and helped KONE to sharpen its position as an innovative company vs its competitors. Proactive demand creation for the service with clients wanting to know more about it – very unusual for the very low-interest elevator maintenance business.