Next best action for pharma is software in which an AI model decides which doctor a drug company should contact next, through which channel and with what message. It also drafts the message, within the rules on what a company may claim about a medicine. The general idea, choosing the most useful next step for each customer, is called next best action. In pharma, using several channels together in this way is also called omnichannel engagement.
For a long time, drug companies reached doctors mainly through the medical representative (MR), the employee who visits clinics. Visits are expensive, and many doctors now have little time for them. So companies also use emails, WhatsApp messages, webinars and articles on a doctor community platform. With so many channels, a company must decide what to send to whom. If it sends everything to everyone, doctors stop reading.
Here is how the model decides. For each doctor, it reads recent data: the doctor's prescriptions, the MR's last visit, which emails the doctor opened and which webinars the doctor attended. For each possible action, it estimates how likely the doctor is to respond. It then picks the action with the best chance. It sends nothing when a doctor has stopped responding.
Imagine a company that sells skin medicines and has 8,000 dermatologists on its list. Every Monday, the model scores each of them. Take one dermatologist in Kolhapur. She has not met the MR for six weeks. She opened two of the last three emails, and both were about results from clinical studies. The model estimates a 54% chance that she will read an email about a new study. It estimates 31% for an MR visit, and 22% for a webinar invitation. So it picks the email.
A language model then drafts the email. It may use only statements from the company's approved claims library. This library is the list of statements about the medicine that the company's medical, legal and regulatory reviewers have already checked against the evidence. In India, the Uniform Code for Pharmaceutical Marketing Practices (UCPMP) requires claims to be accurate and supported by evidence. In the United States, the drug regulator enforces similar rules. A reviewer still approves each new draft before it is sent. This is human in the loop.
If an MR visit had scored highest, the visit and its topic would appear in the MR's field sales software instead. The model runs inside pharma commercial analytics. It is close to an ordinary customer engagement platform, which sends planned series of messages to consumers. The difference is the rules. Every claim about a medicine must be approved before any doctor receives it.