Industrial regulation to protect privacy, intellectual property and proprietary information often restricts data sharing ─ an important prerequisite for developing services in the digital economy. Social media data is publicly available for data mining but requires intensive cleaning and harmonisation before analysis. This paper reveals the process of semantic sensing to convert social network tweets into meaningful insights. Our research question is: how to realise semantic sensing for data innovation? We use design science research to develop an artefact-ontology that collects tweets by pet owners talking about their itchy pet into knowledge graphs, including symptoms, location, breed, timestamp and potential cause and converts them into a thematic map of the regional occurrence of symptoms and potential treatment needs, providing vital information for data innovation. The semantic engine can predict potential causes of itching from the tweet, so a Chatbot may contact the pet owner, inviting them to a veterinary screening. Animal health and pharma companies can use this information to position their services. Our theoretical contribution is a process of semantic sensing, which is a vital part of dynamic capability. Although limited to animal health, the results could be transferred to other contexts.