It’s becoming harder for life sciences to ignore how their audience’s behavior is changing online. More people are using AI tools as their first stop for health-related questions because it’s faster and easier to get a direct answer than searching through a traditional website.
Earlier this month, OpenAI reported that over 230 million people globally use ChatGPT every week for health and wellness related questions.
The risk with this shift is that, on their own, chatbots like ChatGPT, Gemini, or Claude for example, are not designed to meet the strict and unique compliance requirements of the life sciences industry. They can hallucinate and provide an answer that sounds correct but may not be for a specific individual. They use a mix of sources, and they don’t have built-in ways to recognize or escalate serious safety concerns or adverse events.
While these risks call for a more careful approach, it doesn't have to completely hold organizations back from delivering faster, more helpful experiences on their websites. This is where teams are taking a more gradual approach, rolling out AI in ways that match their internal comfort level, regulatory environment, and audience needs.
Platforms like ai12z make this possible by giving organizations tighter control over how a digital assistant behaves. For example, only allowing responses to be sourced from an organization's approved content, preventing it from guessing when information isn’t available, and flagging sensitive interactions that may require human interaction.
We see teams adopting a range of approaches, which we think of as a crawl, walk, run spectrum to involve conversational AI.
The “crawl” stage might position the digital assistant as more of an intelligent guide rather than an answer engine. It elevates the experience beyond a keyword based search, and yet does not directly answer visitors’ questions. For instance, if a patient asks a pharmaceutical website, “How can I get off my antidepressants?", the digital assistant will not summarize an answer. Instead, it scans approved sources and returns a curated list of links with a short summary below each of why it’s the appropriate resource, allowing the user to review the original material themselves.
However, it can still generate an AI response from the organization’s approved content if asked a non-medical question, like company policies, locations, and other general information.
A “walk” stage builds on this by adding an AI overview to search results. In addition to the list of relevant links, the assistant provides a short summary that synthesizes information from those sources. The user still sees and navigates to the original content, but receives a clearer, more accessible response upfront.
A “run” stage represents a full conversational assistant that answers questions directly. There’s no additional risk because the system is still configured to only use approved content. The main difference is how the experience feels to users, and how comfortable an organization is with a more conversational interface.
Instead of listing links or summaries, the assistant uses those sources in the background to generate a complete response. For example, if a user asks, “Is it hard to travel with an insulin pump?”, the assistant scans all the information from available sources and responds with a summary or explanation, such as what to expect at airport security, how to handle screenings, and what precautions to take while traveling.
This flexibility of the ai12z platform gives life sciences organizations more options. They can respond to how people actually engage with brands online today, without being pushed into experiences that might pose a risk to their brand.
If you’d like to hear more about how life sciences organizations are using ai12z, you can get in touch with us using the form below.