Since last year, when Peas In the Pod was founded, we have witnessed the AI revolution unfold before our eyes, reshaping the way we live and work across industries. So we embarked on a journey to explore how the new technologies can be integrated into our workflows. It has been a fascinating learning curve, full of promise, a few hiccups and a lot of “let’s see how this goes" moments. Today, we would like to share some reflections on our experience so far, hoping it resonates with others in the industry who are also navigating this evolving landscape.
What have we tried so far?
1. Transcription and translation: promising but not perfect
One of the first areas we explored was using AI platforms to transcribe and translate recordings of interviews across diverse markets like China, Germany, Austria and UK.
Here is what we found:
AI transcription is cost-effective and time-efficient, especially when managing a large number of interviews.
Accuracy remains a challenge. Brand names, accents and medical language (e.g., molecules, disease names) often lead to errors or unclear transcripts. This means some medical conversations become tricky to follow.
English-to-English transcription works significantly better than translations across languages. For example, note-taking for internal and external calls has been a game-changer.
We are collaborating with an external expert to build an automated workflow to transcribe and translate all interview recordings. We initially thought this would be straightforward, but tailoring a tool to meet our specific needs - like cross-checking for accuracy - has proven more complex than expected. The tool is expected to be ready in the next month or two, and we are eager to see how it performs.
2. Content analysis: enhancing efficiency
AI-powered text-based content analysis tools have existed for a while, but we are aiming to push their utility further by incorporating them into our custom workflow. Here is what we are experimenting with:
Word counts, trend analysis and topic overviews: automating these tasks could speed up topline writing and make identifying keywords, recurring topics and language patterns much more efficient.
Finding language that resonates: tools that can help highlight how HCPs and patients communicate, providing valuable insights into tone and phrasing.
However, there is a limitation: AI tools are constrained by the number of tokens (characters or words) they can process in a single session, which means long transcripts or in-depth conversations may require manual intervention.
3. Building better patient profiles
In the realm of content analysis, one area where AI has truly shone is in supporting the creation of patient profiles.
AI has streamlined our process, ensuring no detail - no matter how small - is overlooked.
It helps us build richer, more encompassing profiles by synthesising everything respondents share.
We also see great potential in using AI for market and customer segmentation, making these processes far more efficient and comprehensive.
What's next?
1. Sentiment analysis
We are planning to integrate sentiment analysis into our custom tool to detect the mood and tone of conversations in text-based data.
Additionally, we are exploring the potential of applying sentiment analysis to audio and video recordings, helping us gain deeper insights into their underlying emotions.
Why is this exciting?
The idea that AI could identify tone of voice, hesitations, microexpressions or mood changes during a conversation feels groundbreaking.
Why are we cautious?
We’re unsure how accurate and reliable these tools are. Can AI truly differentiate between hesitation from uncertainty versus thoughtfulness? Will it confuse sentiments or misinterpret cultural nuances? These questions remain unanswered for now.
2. Social Media Listening
This is where AI really complements a classic methodology. We are exploring how social media listening with AI could integrate into our approach:
Collecting unprompted data from global audiences.
Spotting emerging trends in real time, providing a starting point for deeper qualitative investigations.
Incorporating sentiment analysis to identify mood and detect nuanced feedback.
It is faster, cost-effective, and offers a broader reach, a great way to connect with larger and more diverse audiences.
How do we feel overall?
While we are excited about the potential of AI, we are also aware that translating our vision into reality is not straightforward. There is still so much to learn, test and refine. And to be honest, it is quite overwhelming!
One important consideration we are paying close attention to is data security and privacy concerns. With sensitive pharmaceutical data and patient insights involved, it is essential to ensure compliance with data protection regulations such as GDPR. Choosing platforms that prioritise encryption and secure storage is non-negotiable for us, and we are committed to maintaining the confidentiality of all data throughout our AI integration process.
But here is what we know for sure:
AI is not replacing human researchers anytime soon, but it is proving to be a valuable partner in streamlining processes and augmenting insights.
The key is staying curious, experimenting and learning what works (and what does not) for our field.
If you are also navigating the integration of AI into market research, we would love to hear your thoughts and experiences! What has worked for you? Where are you seeing challenges?
Peas In the Pod is a global insights, strategy and communication agency specialized in the life science business. Our mission is to contribute to the creation of a world where the patient, the prescriber and the pharma organization - the Peas - are in perfect alignment, listening to each other, thus fostering understanding and collaboration.
*Connecting the dots between the patient, the prescriber and the pharma industry*