GPT-assisted Insight Generation and Data Navigation


Project Overview

Prospection has looked to transform the way people in pharmaceutical business teams manage their products with the introduction of a GPT-powered assistant to its patient-centric-intelligence-powered platform. This assistant, designed with a deep commitment to explainable AI as a guiding principle, removes capacity, experience and expertise barriers from teams looking to use patient-centric intelligence to achieve their objectives.

Patient-centric intelligence is a multidimensional approach in healthcare that places patients at the forefront of decision-making processes. As a pioneer in patient-centric intelligence Prospection has built a capability for ingesting, analyzing, and interpreting patient claims data, enabling the provision of products and services that can be used to improve healthcare outcomes, enhance patient experiences, and optimize commercial performance.

The introduction of Prospection’s GPT-powered solution has immediate impact on the launch and management of pharmaceutical products, but goes beyond that to become a key lever in accelerating an organisation’s ability to navigate further along the data maturity model. This model labels initial use of data as “descriptive” (ie what is happening), moving then to “diagnostic” (ie why something is happening), next to “predictive” (anticipating future trends and states) and finally to “prescriptive” (where best course of action are recommended based on the available data).




Prospection Product and Engineering Team have done the development. However our analytical consultants have helped train the LLM.

Project Brief

Implement an LLM/GPT-powered service as a means to improve the user experience and expand access to data-driven insights across life sciences business teams. Make it easy to get answers to the key business questions that teams have, give teams the ability to explore and refine insights, help upskill the analytics capabilities of business users.

Project Need

The pharmaceutical industry has long been a champion of a data-driven approach to success and though in recent years its use of patient-level data has expanded, this kind of intelligence remains under-utilized as a means to improving patient outcomes and optimizing brand performance.

This under-utilization can be attributed to a few main reasons:
- Business users - brand managers, marketing managers and commercial managers - do not always have the analytics experience required to make use of business and insights reporting tools, which are notoriously complicated and often require BI specialists that can use and get answers for business-user questions.
- BI specialists are a limited resource and are often shared across multiple brand teams. This can mean their ability to deep dive into specific clinical areas is constrained, thereby limiting their ability to operate as deeply connected business partners to their colleagues.
- Sales data has been the conventional view of brand performance, and is relatively simple. While patient-claims data has been used in some parts of the pharmaceutical business, like medical affairs, its use as a commercial tool has not been widely adopted.

User Experience

One of the key UX principles was to augment the existing experience - to help democratise data and analytics - without impacting the experience of our power users.
To do this we designed a GPT type interface as an abstraction layer on top of the existing UI. With this approach, natural language is used to drive the UI, which greatly reduces the learning curve for new or infrequent users. This approach also has the benefit of helping to educate users on how to use the analytics platform.
Another important principle we used was “recognition over recall”. Many prompt-based interfaces can be very intimidating as they are just an empty input box. This relies on the user having to come up with (recall) a useful prompt. To overcome this, we provide example prompts that the user could simply select (recognition) to start experiencing the capabilities of the GPT type interface.
We recognised early on, through user testing, that there were different use cases for the new interface. Rather than design a single UI that attempted to cater for all use cases we designed different modes. This allowed us to optimise different GPT modes for different use cases, giving users the best possible experience.
Finally one of our foundational principles behind AI, focused on transparency or as it's becoming known in industry "X-AI"
meaning explainable AI. In the video provided you'll also see a"Explain how you did that" link this was to remove the AI "black box"

Project Marketing

The GPT assistant feature is part of a repositioning of the entire platform, moving away from being broadly targeted as a data service used by BI teams to being a patient-centric intelligence platform that helps business users self-serve insights on demand. This is critical in supporting Prospection's journey from a services-led business to embracing a product-led approach, and becoming a true SaaS product.

Project Privacy

Prospection’s platform operates under strict privacy and data usage policies, and adheres to legal and ethical standards.

We prioritize patient data privacy and adhere to strict privacy regulations and best practices. When querying patient claims data, GPT analyzes de-identified information, ensuring the anonymity and privacy of individual patients. No personally identifiable information is utilized or exposed during the querying process.

The use of GPT in querying patient claims data is designed with utmost consideration for data security. Prospection ensures that an organization's data remains protected and confidential throughout the process. No data from any organization is shared or exposed during the querying process.

Using GPT for querying patient claims data does not compromise an organization's proprietary data or competitive insights. Prospection is designed to operate independently, analyzing publicly available or commercially accessible data sources to answer specific questions and generate insights. An organization's proprietary data and competitive advantage remain protected and secure within their own systems.

The AI and GPT has the power to reshape our world, as there is endless opportunity to create new possibilities based on data, human need and AI generative methods. What will stand out in Digital - AI - GPT are those projects that create a seamless experience and make the ordinary extraordinary.
More Details