A senior director recently told me, ‘I’m afraid my team is falling behind.’

That fear came to life during my recent AI Masterclass with Life Sciences Access Academy, where attendees lit up the chat with questions about how to make AI work for MSLs. The divide between teams embracing AI and those hesitant to adopt it is growing fast. The ones who aren’t keeping up risk getting left behind.

In the spirit of being an innovative AI user, I put the chat history from the masterclass into AI to identify the top questions. The results? Three major themes emerged, highlighting the exact challenges and opportunities MSLs are facing as they integrate AI into their work.

Here are the top three AI for MSLs questions everyone asked and the answers that will help you and your team stay ahead.

1. Which AI platforms are best?
2. How do I get better AI outputs?
3. How do I tell if AI is hallucinating?

Before you get the answers to the top 3 AI for MSLs questions, get a high level overview of what was happening in the chat:

1️⃣ Which AI Platforms Are Best?

👉 “Hi all – interested to hear WHICH AI platform(s) you use – and when do you use one over another?”
👉 “Is there are list ranking currently industry used AIs, i.e. is our company AI good enough?”
👉 “Does the free version of ChatGPT work for this?”
👉 “How do you decide when to use external GBT (optimal system) and internal systems (sub-par)”

The Answer: Follow Your Company’s Policies and Choose the Right Tool for the Job

When it comes to AI for MSLs, your company’s policies should guide your choices. Many organizations have strict rules about which tools are permitted due to compliance concerns and data confidentiality. Even if external tools like ChatGPT or NotebookLM are tempting, always ensure you’re following company-approved processes.

Once you understand your company’s AI policies, choose the right tool for the specific task:

• ChatGPT: Great for pre-meeting planning and summarizing documents, but requires careful prompting and compliance checks.
• NotebookLM: Ideal for summarizing and verifying large volumes of documents—especially useful for dense scientific literature.
• SciSpace: Built for deep analysis of scientific papers and identifying research trends.
• Custom GPTs: Perfect for repeatable tasks like KOL profiling but require significant setup and customization (check out Vivek Mukhatyar’s custom GPTs built for MSLs here).

Takeaway: Start by understanding your company’s policies. Then, based on what’s allowed, select the tool that aligns with the job you need to accomplish. This ensures you’re using AI effectively and responsibly.

2️⃣ How Do I Get Better AI Outputs?

👉 “I find no matter how I respond to the insight, trying to edit or improve the quality, there is a ceiling for the grading of the insight. The AI seems hyper critical.”
👉 “I’ve used AI to analyse insights and it seems that it always needs a human sense check but there are some good top-line outputs”
👉 “We are using an external agency to run our AI reports with insights but even with lengthy, detailed insights we find the output is still quite generic”

The Answer: Better Inputs = Better Outputs

AI is only as good as the data and prompts you feed it. If your inputs are vague, you’ll get vague results. And if the data going into AI is incomplete or low quality, no amount of prompting will make the output meaningful.

To get high-quality AI outputs, focus on two things:

1. Ensure the Data Is High-Quality

– AI can’t fix bad insights. If the data lacks context, the why, or implications, AI will only amplify those gaps.
– Before using AI for insights analysis, make sure the raw insights are complete.
– Garbage in = garbage out.

2. Write Better Prompts

– The more specific and context-rich your prompts are, the better AI’s response will be.
– AI is not a mind reader. It needs clear instructions.

Get more tips on how to craft better prompts in my AI guide for creativity and productivity 👇.

Examples of Effective Prompts from the Masterclass:

• Insight Analysis (internal tools only!): “You are an expert in analyzing msl insights and identifying trends and decision grade insights. your task is to analyze 100 insights and create a report of the major trends. here are the insights:”
• Pre-Meeting Prep: “You are an experienced senior MSL who is an expert at building relationships with KOLs. Your task is to help me pre-meeting plan for my upcoming meeting with a top KOL. This is my first meeting, and my goals are to build the relationship, understand her needs and how I can add value, and get another meeting. I need your help with setting meeting objectives, developing questions to help me build the relationship and uncover her needs, potential resources/data the KOL may be interested in, and setting the agenda. I also need your help with how to identify ways to open the conversation (such as reviewing recent publications). Please also suggest additional things I am not considering. Keep your answer brief and concise.”
• What Your KOL Values: “You are an experienced senior MSL and an expert in identifying what different KOLs value to help build long-term relationships. I need you to help me understand how I can add value to my upcoming meeting with a Key Opinion Leader (KOL). Here’s what I know so far about the KOL:”

Takeaway:

The quality of the input, both the data and the prompt, directly determines the usefulness of AI’s output. AI won’t fix bad insights, but it can amplify high-quality inputs to generate meaningful, strategic results. Train MSLs to capture better insights and craft precise prompts, and you’ll immediately see better AI-driven outputs.

3️⃣ How Do I Tell If AI Is Hallucinating?

👉 “Do you have any advice or tips for how to quickly identify hallucinations? Especially in analysis of literature, it seems like it could take just as long to review the answers as it would be to read a paper in the first place.”
👉 “Hallucinations are tricky! But there are methods available to minimize these :)”
👉 “How much of an issue are hallucinations with this service?”

The Answer: Always Keep a Human in the Loop

AI hallucinations occur when the tool generates false, misleading, or entirely fabricated information. This is especially dangerous in Medical Affairs, where accurate data is critical.

A big part of using AI effectively is having the right mindset about what AI can and cannot do. AI is not here to replace MSLs. It is simply a tool to assist, streamline, and enhance your work. Expecting AI to do everything on its own will lead to frustration and mistakes. Instead, think of AI as an over-enthusiastic intern. It can process information quickly, but it still needs oversight.

Here’s how to minimize the risks of AI hallucinations while setting realistic expectations for how to use it:

1. Understand the Tool’s Capabilities:

– Know what AI is good at (summarizing, pattern recognition, structuring information).
– Know what AI is not good at (math, fact-checking, original thinking, strategic decision-making).
– AI is a helper. It doesn’t replace critical thinking.

2. Ask for Sources:
– When using AI for literature analysis, always ask where the information is coming from.
– Example: “What is the source of this data?” Some tools, like SciSpace or NotebookLM are good at showing you where it pulled the info from.
– If AI doesn’t provide a direct citation or gives a vague answer, don’t trust it blindly. Verify it.

3. Cross-Check with Original Documents:

– Always compare AI summaries or insights with the original papers or data.
– AI might miss key details or overemphasize certain aspects, so manual review is essential.

4. Fact-Check Key Points:

– If AI provides a recommendation or highlights a trend, validate those points manually before using them in strategic discussions.
– AI can spot patterns, but it doesn’t always understand what’s actually important in a medical or business context.

Key Insight from the Chat:

“AI is an Assistance, the quality of the onboarding process of this assistant determines the quality and productivity.”

Takeaway: AI is a powerful assistant, but that’s all it is: an assistant. It can help MSLs work faster and smarter, but it still requires human expertise to ensure accuracy and strategic relevance. If you think of AI as a tool to support your thinking, not replace it, you’ll get far more value out of it.

Conclusions: AI for MSLs Don’t Let Your Team Fall Behind

AI is here to stay, and the divide between teams embracing it and those falling behind is growing fast. To stay ahead:

• Understand your company’s policies: Always start by following your organization’s AI guidelines.
• Use the right tool for the job: Once you know what’s allowed, choose the tool that aligns with your specific needs.
• Keep a human in the loop: Validate AI outputs, cross-check sources, and ensure accuracy in every step of the process.

📢 Want to master AI for Medical Affairs? Check out my AI for MSL Excellence Training, where I cover:

• Best practices to get AI to work for you
• Prompt engineering for better results
• Recognizing what hallucinations look like

What’s been your experience with AI so far? Are you using it effectively, or are you hesitant to dive in? Share with me.