pfizer

Problem / Challenge

Drug information can be difficult to understand, often filled with complex medical terminology that can confuse rather than reassure a patient. Pfizer recognised this problem and wanted to explore how Voice could help Australians with easy-to-comprehend information around their treatments

Process

I conducted a series of discovery workshops with Pfizer medical experts and executives. Through a series of workshop exercises, we mapped key customer journeys and business processes to identify moments of disproportionate impact. These moments unlocked opportunities conversational AI could be used to improve the experience for the patient. 

From the sessions, we learnt that on the patient side, taking newly developed drugs could have unexpected side effects and they felt a lack of support when it came to addressing these. On the business side, Pfizer has an expensive inbound call cost in addressing these issues. 

Taking a data-driven approach to build out our language models, I mined call centre data and logs to identify common patterns for contact before creating a framework to help manage and retrieve unmapped utterances through user testing.

After research with Pfizer physicians, we understood that conversational AI wouldn’t be able to help support everything a user needed along their journey. We designed mechanisms that used AI to recognise when we couldn’t provide information on a question and organised a handover to a trained medical practitioner through the experience, as well as an integration to their CRM.

We viewed the assistant as a virtual nurse. They can’t offer any medical advice but can guide you through treatment and remind you of the directions associated with the drug you are taking.

Outcomes

We solved a key business problem for Pfizer, through the reduction of call centre volume. Patients experienced a reduction in call centre wait times, with the most common queries being answered through voice.