Artificial intelligence (AI) has moved far beyond being a buzzword; it’s now woven into the fabric of our everyday routines. From streamlining communication and simplifying email management to boosting research capabilities and guiding smarter decisions, AI tools are built to help us work faster, efficiently, and smarter.
That said, how we use AI matters just as much as the technology itself. At UBC, we believe that exploring AI must be done responsibly, always aligning with privacy and compliance standards.
This blog is part of a series on AI in pharmacovigilance (PV). Each post takes a PV-centric perspective, looking at how AI’s diverse capabilities could inspire innovation and reshape the way we work. These are conceptual explorations; it does not utilize UBC clients’ use cases or deliverables. Instead, the use cases are meant to spark discussion, share insights, and open possibilities for the future of PV.
With so many AI models available, each offering distinct strengths, the real challenge is knowing which one to pick. Think of them as a team of smart assistants: some excel at quick, simple tasks, others shine in deep analysis and problem-solving, and a few are purpose-built for coding or tailored to organizational needs.
By recognizing these different “personalities”, we can match the right tool to the right task and unlock new ways to optimize our workflows.
General-Purpose AI Models Designed for Performance & Deep Reasoning
A. GPT-4o – Optimized for Speed & Everyday Productivity
Open AI has developed several deep learning AI models since the last years; the first breakthrough took shape of a Generative Pre-trained Transformer (GPT) version 3.5 and was followed by its brothers and sisters. As a result, we now have access to GPT-4o, which is part of the GPT-4 family but optimized for speed, efficiency, and lower cost.
Just spent an hour on taking training notes? Feed them to GPT-4o for a quick bullet-point summary to get organized! Need to share key takeaways with colleagues who missed the training? Ask for a brief summary! How about creating a chatbot with your notes? It can answer your colleagues’ questions about the training super handy!
Tackling dense meeting minutes filled with critical PV details? Upload the 20-page document to GPT-4o and watch it transform into a sleek bullet summary in seconds! Need to update leadership? Get a concise “5 key risks” section instantly. Want to align your team? A polished 1-slide summary is just a request away.
Work smarter, not harder! Use GPT-4o for everyday tasks like summarizing articles, drafting quick emails, or preparing meeting notes. It makes summarizing effortless and fun!
B. GPT-5 – Advanced Reasoning for Complex Workflows
Get ready to dive into the thrilling world of regulatory compliance with GPT-5 as your trusty sidekick!
GPT-5 innovates from previous OpenAI models by being smart about how to answer, such as determining if it should provide a simple answer or go into a deep analysis for you. For example, you could unleash the power of GPT-5 for your regulatory adventure by benefiting from its advanced reasoning capabilities!
Start by defining your mission ~ perhaps focusing on “Global Expedited Reporting Conflicts” when regulations are conflicting, and you need to dive deep into these regulations. And why not share the new insights you gained in your new white paper? GPT-5 can help you set clear objectives, strategy brief, or engaging heat map, depending on your needs. Let’s see what you would need:
- Know Your Audience: From PV leadership to auditors, tailor your analysis just right!
- Gather Your Resources: Collect key documents like ICH guidelines and regional regulations. GPT-5 can process hundreds of pages, and build a comprehensive knowledge base just for you.
- Analyze & Align: Let GPT-5 help you navigate the complexities of drugs, devices, and combination products. It’ll highlight differences in reporting definitions and compliance risks, creating a clear conflict matrix.
- Scenario Modeling: Play with “What If?” situations! Imagine handling a case with partial unblinding, and GPT-5 will simulate predicted workflows and report timelines, flagging potential conflicts and offering mitigations.
- Craft Your Whitepaper: With all that insight, GPT-5 helps you draft a polished whitepaper, including executive summary, regulatory comparison tables, and eye-catching visuals. Then, review, add your touch, and let GPT-5 refine it to perfection.
- Deliver Your Masterpiece: Polished whitepapers, handouts, slide decks, and customized versions can then be delivered to different audiences. Don’t forget to embed it into your organization’s knowledge system for easy access later on!
C. Claude 3.7 Sonnet – Clear & Budget-Friendly
Anthropic has worked hard to propose competitive AI models to the one offered by Open AI. Anthropic proposes two main models with their own interesting properties, the first one, Claude 3.7 Sonnet can be the best ally of your team, for example supporting knowledge acquisition. Get ready to onboard your new entry level PV member with a fun and dynamic 30-minute training session! But how you may ask; simple, gather all the relevant SOP, work instructions, and other supporting documents, then feed them to Sonnet. Next, start asking your new AI trainer to create an energizing agenda that sets the tone and lightens the mood with a quick, funny quiz on PV terminology to get everyone laughing and learning. Ensure your trainee masters key objectives such as case triage essentials, timeline assessment strategies, SAE follow-up expectations, and handling device events.
Dive into real-world scenarios by setting up sample cases for prioritizing urgency or navigating SAE follow-up questions perfect for sparking discussions and teamwork! Wrap it up with a recap to reinforce key points and throw in a rapid-fire Q&A challenge to test their knowledge against the clock. Plus, use Claude Sonnet to effortlessly generate your agenda, objectives, and exercises, making your role as the trainer a breeze! With this lively approach, your new hire will leave excited, prepared, and ready to rock their role in the PV team.
D. Claude 4.0 / 4.1 Opus – Deep, Thoughtful & Strategic
Imagine tackling a mountain of essential documents like your Risk Management Plan (RMP), Safety Management Plan (SMP), and Development Safety Update Reports (DSUR). Sounds daunting? Fear not! With Claude Opus, you’ll breeze through this overwhelming task. Just upload your files, and watch as Claude crafts a cohesive understanding of safety concerns, pulling together key events into an engaging narrative. It’ll spot inconsistencies, highlight recurring themes, and mark risks that need escalation, ensuring everyone knows their role in addressing the safety profile.
Introducing Claude Opus, your ultimate “AI SOP Librarian” for PV! This super-smart assistant could be storing and navigating your library of procedures from SOPs to regulatory guidelines making finding the right info a snap. Need the follow-up timeline for device SAEs in the EU? Just ask! Claude could dive into guidelines, pulls exact requirements, and explains critical differences between device and drug reporting, all while linking back to original documents for audit-ready accuracy. Claude could also resolve conflicting procedures seamlessly, ensuring consistency across programs. Whether it’s compliance checking or procedural impact analysis, Claude Opus makes your PV life easier and more efficient. Who knew consolidating documents and learning could be this fun?
Coding & Technical Support AI: Powering Automation in PV
A. GitHub Copilot
Copilot is the AI made by Microsoft and primarily known to integrate across all Microsoft 365 applications, which we will discuss later, but it also lives inside your favorite code editor as GitHub Copilot as your magical coding sidekick that supercharges automation in PV!
Imagine exporting hundreds of Argus cases into Excel and then struggling to reconcile them with your clinical data using an electronic data capture (EDC) extract or a QC tracker. You’re trying to spot missing seriousness or causality assessments, mismatched expectedness, or incomplete fields; matching and verifying quickly becomes overwhelming. That’s where Copilot steps in. Simply describe what you need: “Compare these two files by Case ID, flag missing or inconsistent safety fields, and generate an exception log”. Copilot generates the Python script for you; behind the scenes, the script reads your files, stores them in clean data frames, highlights gaps and mismatches, and produces a clear report you can share with your team.
But it doesn’t stop there! Copilot can help you write SQL queries to check pre-write quality control rules for automated audits, create dashboards for trend monitoring, and even flag anomalies like a pro. What once took weeks of coding is now reduced to days, turning automation into a practical, repeatable workflow for your developer or even an empowering new skill for you!
B. Windsurf
Get ready to ride the wave of efficiency with Windsurf during your PV Automation Sprint! Imagine leading your team to integrate Argus data into a PowerBI dashboard, automate quality control rules, and develop an E2B transformation tool. Sounds exciting, right? That’s where Windsurf, your trusty project management tool, steps in, transforming chaos into streamlined success. Windsurf tracks all your tasks, so you can see who’s doing what at a glance; no more frantic searches. As your team navigates the automation seas, it generates dependency diagrams that show how each task connects. Need to prioritize? Windsurf suggests which tasks to tackle first! For the development team, it also organizes GitHub issues, helping you focus on what matters most, and produces comprehensive engineering documentation for easy future references. With Windsurf by your side, your PV automation team can surf the waves of productivity, making project organization not just effective but fun! Learning to leverage this tool will turn your sprint into an adventure rather than a chore.
Business-Focused AI Solutions Tailored for Efficiency
A. Microsoft 365 Copilot – The Game Changer for Microsoft-Based Companies
Microsoft 365 Copilot is an AI-powered productivity assistant built directly into Word, Excel, Outlook, Teams, OneNote, and PowerPoint, transforming everyday tools into intelligent collaborators. Unlike stand-alone chatbots, Copilot understands your organizational context emails, files, meetings, and chats. This allows Copilot to be more effective in drafting emails, building documents or presentations, and analyzing data. It can do so with a prompt and a file input to transform, for example, a presentation into a descriptive report.
For PV teams, Copilot acts as great support to better handle your emails, summarizing a long email thread you lost track and then asking for a draft answer including the decision you have taken on the ask from your team. Or ask it to take meeting minutes during the weekly catch-up with your team, noting action items so you can focus on the discussion and not on taking notes. It could help you find the correct wording when you need to update an SOP. This tool doesn’t replace the years of expertise your team has; rather, it amplifies the expertise by removing administrative workload, speeding up decision-making, and enabling teams to focus on strategy and high-value oversight.
B. Retrieval-Augmented Generation (RAG)
Imagine stepping into the dynamic world of pharmacovigilance as a new safety scientist with a pressing question: “What’s the follow-up timeline for device malfunctions in the EU?” Instead of hunting through a maze of documents, just turn to RAG to build your trusty knowledge-based assistant! Simply ask, and RAG retrieves the exact SOP paragraph you need to answer your question; you no longer need to dig into and search countless documents. But there’s more! Using RAG, it is possible to optimize the summary of key points, in the format usually used by your company, delivering a clear response with corresponding citations for audit-ready credibility.
But how does this work, you might ask? Well, remember the SOP librarian we mentioned earlier? It is based on RAG – this is not an AI model but rather a technic to adapt a trained model on a more specific domain of expertise. Whether you’re onboarding, ensuring SOP alignment, or need quick answers, RAG works well because it is connected to your data. It turns your onboarding experience into an interactive journey, saving you time and building a solid foundation for PV trainees.
C. Custom AI Models
Welcome to the future of PV, where automation and AI work together to create a smarter, faster, and more compliant safety ecosystem. The automation layer drives the operational workflow, automatically interpreting device event codes, proposing seriousness levels, flagging missing E2B(R3) fields, identifying unexpected terms or serious injury criteria, triggering AOSE or 7-day reporting alerts, and generating structured outputs such as MedWatch drafts, CIOMS narratives, and clean missing-field checklists. These rule-based actions are executed in seconds via automation, eliminating repetitive manual steps.
The AI layer, powered by custom PV models, brings intelligence and reasoning to the process. Thing of custom AI models as simplified GPT models; instead of being trained in immense datasets, they will be trained on large internal datasets, relevant to ensure that these models support the specific task we would like to give it.
For example, by interpreting clinical language in AE reports, AI suggests tailored follow-up questions, generates narrative-style summaries, and integrates outputs seamlessly into emails, documents, presentations, trackers, and stakeholder communications. Together, automation accelerates the workflow, while AI enhances understanding, drafting, and decision support creating a fully integrated PV environment that transforms complex daily tasks into a streamlined, intelligent process. The take home message is that you might not need to have a Ferrari to cross the street.
The “best” tool isn’t universal; it’s all about picking the right one for the job at hand.
| Model/Tool | Strengths | Generic Example | PV Example |
| GPT-4o | Fast, cost-effective | Summarizing articles | Summarizing meeting minutes or client updates |
| GPT-5 | Deep reasoning | Whitepapers | Multi-guideline global reporting strategy |
| Claude Sonnet | Simple, clear | Drafting emails | Training outlines, checklists |
| Claude Opus | Long memory, strategic | Business roadmaps | Reviewing full PV archives, guidelines, DSUR history |
| GitHub Copilot | Coding efficiency | Autocomplete code | Automating Argus recon & QC checks |
| Windsurf | Technical PM | Sprint planning | Managing PV automation pipelines |
| RAG | Internal knowledge retrieval | HR queries | SOP lookups, program-specific reporting checks |
| Custom AI Models | Compliance-focused | Fraud detection | DEC interpretation, QC auto-check, SAE/SADE classification |
The truth is, there isn’t one single “best” model.
Yes, you heard that right!
The secret sauce is finding the perfect model for your task. AI is here to make life easier and more efficient, so choose the one that feels like your ideal match. Sometimes it means using a simpler approach than using the latest, biggest model.
We hope you have gathered some great insights on models like GPT, Claude, and Copilot. Now go forth and unleash these innovative tools to supercharge your work and creativity!
Read other UBC PV AI Blogs:
Putting a Seatbelt on AI: How Risk-Based Integration is Transforming Drug Safety – UBC
What Place for Humans in a PV World Integrating AI?
About UBC
United BioSource LLC (UBC) is the leading provider of evidence development solutions with expertise in uniting evidence and access. UBC helps biopharma mitigate risk, address product hurdles, and demonstrate safety, efficacy, and value under real-world conditions. UBC leads the market in providing integrated, comprehensive clinical, safety, and commercialization services and is uniquely positioned to seamlessly integrate best-in-class services throughout the lifecycle of a product.
About the Author
Poonam Pathak, Team Manager, Global Case Processing
Poonam Pathak joined UBC in 2021 and currently leads global teams responsible for pharmacovigilance (PV) case processing operations across multiple complex drug and device programs. In her current role, Ms. Pathak oversees end-to-end case processing, regulatory submissions, medical review workflows, vendor oversight, and cross-functional coordination with sponsors and CRO partners. She is actively involved in process and operations setup, project governance, and continuous improvement initiatives to enhance quality, compliance, and efficiency across global PV activities.
Ms. Pathak also serves as a PV process lead for multi-study environments, supporting Argus configuration alignment, Safety Management Plans, AERP updates, device reporting requirements, and workflow optimization, including emerging PV automation and dashboarding capabilities.
Prior to joining UBC, Ms. Pathak held progressive PV leadership roles overseeing case processing, narrative quality, team training, compliance tracking, and cross-regional regulatory operations. She has extensive hands-on experience in managing global safety deliverables, writing and reviewing SOPs and work instructions, collaborating with regulatory partners, and ensuring operational excellence across high-volume programs.
Ms. Pathak is currently pursuing an MBA with a focus on strategy and leadership and has completed several professional trainings in global PV operations, regulatory reporting, quality oversight, device safety, and Argus database management.
