To Build AI Maturity, Start With Quick Wins

Rohit Gupta
Vice president of analytics strategy
and transformation
Beghou Consulting

Despite the significant hype around artificial intelligence in the pharmaceutical industry, many commercial teams still operate at a low level of AI maturity. This is understandable given that these are emerging technologies. However, the ultimate goal of pharma companies’ AI initiatives should be to develop enterprise maturity. A company maximizing its AI capabilities would have tools fully embedded in its operations, which would be trusted by users across the organization–from the field sales force to the C-Suite. These tools would not only provide insights but also drive actions (e.g., next best engagement actions for the field sales force).

Building up to this level of maturity is a significant lift, and for a company starting at a low level of maturity, it can sound like a daunting proposition. To succeed, pharma companies must be methodical and strategic instead of overly aggressive in their pursuit of AI maturity. This approach includes:

• Understanding and figuring out ways to navigate around the shortfalls and risks of AI.
• Creating an achievable AI roadmap for the organization and managing change effectively along that roadmap.

Most importantly, to build long-term AI maturity, a company should pursue focused AI projects that generate quick wins over complex, large-scale initiatives. In our work with pharma companies–especially innovative emerging companies–we have seen that this approach helps companies generate results from AI in the short term while building enterprise maturity that eventually enables large-scale AI deployments.

Fit-for-purpose vs. fit-for-everything

With a bewildering array of AI tools on the market and the many risks associated with the technology (related to data privacy, hallucination and more), companies need to be targeted in their deployment for commercial operations. Attempting to execute large-scale AI projects too quickly can lead to unnecessary complexity and delay the impact a company sees.

Amish Dhanani
Associate principal
Beghou Consulting

That’s why we recommend pharma companies pursue “fit-for-purpose” instead of “fit-for-everything” solutions. While it’s tempting to think about potential future applications of an AI solution, some (or all) of these future applications will never come to fruition. Instead, we urge the pharma companies we work with to focus on a specific business challenge when pursuing a project. Here’s how we recommend honing an AI project down to its essential elements to maximize precision and impact:

1. Define the business challenge. Commercial leaders must start by understanding the business challenge they hope to overcome using AI. That business challenge must be specific (things like improving the response rate to digital marketing materials among target hematologists, instead of high-level goals like growing market share). By homing in on a specific business challenge, commercial leaders can determine the best way to leverage AI in developing a solution.
2. Understand the available tools. Pharmaceutical commercial teams should be exploring the available AI tools and assessing their fit for their potential applications. This knowledge of the tool landscape will ensure that commercial leaders make sound decisions instead of being pulled in various directions by vendor sales pitches. Commercial teams should also be flexible in their use of tools. New tools will emerge as AI matures, and companies should position themselves to shift to better solutions when they come on the market.
3. Is generative AI the solution or a distraction? Generative AI has great potential to unlock insights and speed delivery of information across a pharma commercial organization. Everyone is talking about it and wants to find a way to use it, but it’s not the right fit for every project. In some cases, more established forms of AI, such as natural language processing or robotic process automation, make the most sense. For example, if a company is trying to provide structured answers and a single source of truth, methods like RPA will work better than generative AI. On the other hand, if a company wants to develop an engine that provides answers to an array of unanticipated questions or guidance on complex, decision-driven tasks, generative AI tools are likely a better fit. The bottom line is that commercial leaders shouldn’t let the desire for cutting-edge solutions disrupt their focus on pressing business needs.
4. Be agile and keep making progress. There’s an inherent level of experimentation involved in the deployment of AI solutions today, and nothing is static. There will be successes and failures in AI deployments, so companies must be agile and responsive to those failures, tweaking and optimizing continually as they build out their solutions. At this point, we recommend companies view AI deployments as “pilots” to encourage this ongoing experimentation. This approach will help a company develop a proven and robust solution.

Here are two examples of how this fit-for-purpose approach to AI deployments played out with pharma companies we partnered with.

Case study: Executing AI-powered precision targeting

An oncology company we worked with needed to create a precise target list of high-value health care providers (HCPs) and key opinion leaders (KOLs) ahead of its product launch. The company sought to move beyond lagging data and capture shifts in HCP and KOL sentiments and influence. After all, HCP and KOLs’ views of therapy can change frequently, and someone who is largely unknown one day can become a minor celebrity the next thanks to industry accolades or social media activity.

To build an AI engine that tracked HCP and KOL sentiment and influence, we:

• Combined proprietary data with data from public databases (e.g., society and conference websites, medical literature repositories, social media, health care organization websites).
• Created structured data out of unstructured data (e.g., text from social media, website copy) using natural language processing tools, including large language models and BERT models.
• Used these AI tools to perform aspect-based sentiment analysis and generate insights on HCP and KOL sentiment and reputation.

The AI-powered solution we built helped this company dynamically track shifts in sentiment and reputation and improve its HCP and KOL targeting efforts.

Case study: Improving clinical trial recruitment with NLP

A rare disease company sought to improve its recruitment and retention of clinical trial participants. To gain insight into the field team’s interactions with the company’s contract research organization and HCPs, the clinical team had to sort through notes in a variety of formats (e.g., digital, handwritten, etc.) and with a mix of information. Combing through these materials and uncovering insights was manual and time-consuming, and this inefficient process delayed insight-generation and limited the ability of the company’s clinical team to respond promptly to issues with recruitment and retention.

To address this company’s challenge, we built a natural language processing model that tagged and categorized notes in an intelligent way, accurately interpreting typos, acronyms, and indirect references. Using this NLP model, the company was able to send insights to relevant stakeholders and prompt timely action related to clinical trial recruitment and retention.

Quick wins enable large-scale AI initiatives

The takeaway from these stories is that companies can realize meaningful impact relatively quickly by building targeted, fit-for-purpose AI solutions instead of pursuing over-engineered initiatives. The key to building long-term AI maturity is managing the scale of AI deployments in the short-term. By deploying fit-for-purpose solutions designed to address a specific business challenge, companies can generate impact from the technology that helps them build organizational expertise–and creates a foundation for larger-scale AI initiatives.

Every pharma company has great ambition when it comes to AI today, and they should embrace that ambition. For pharma commercial teams, it’s not a choice between small AI and large AI. The former enables the latter while moving too quickly toward the latter impedes organizational maturity development.

The quick wins a company generates with this focused approach will fuel long-term AI success and commercial advantage.

About the authors

Rohit Gupta is vice president of analytics strategy and transformation at Beghou Consulting. Amish Dhanani is an associate principal at Beghou Consulting.