Emerging AI Proof of Concepts for Accelerating Submission Timelines

Generative AI (GenAI) is drawing significant interest across the pharma industry, prompting leaders to rethink traditional value drivers for their business goals. Large biopharmas are heavily investing in GenAI-driven proof of concepts (POCs) for varying regulatory use cases. Many are betting on long-term returns with reduced dossier authoring timelines and enhanced data-driven decision-making on submission strategy.

However, navigating this new territory comes with complex challenges around non-deterministic responses, such as different answers to the same question, inadequate accuracy, non-granular traceability, security, GxP compliance, and upskilling costs. The current wave of GenAI experimentation will determine whether it matures into an industry-disrupting innovation or remains a mere POC. In this wave, large biopharmas are exploring the value of GenAI for automating regulatory intelligence generation, streamlining document translations, leveraging responses in health-authority (HA) queries, and accelerating dossier authoring.

Automating regulatory intelligence

Industry leaders have shown interest in automating regulatory intelligence. They are using GenAI to summarize large volumes of unstructured HA guidance documents and internal local intelligence data. This will help to:

• Automate the summarization, categorization, and tagging of regulatory intelligence documents.
• Leverage large language model (LLMs) chatbots to analyze data sets, identify patterns, and generate insights for easier consumption.

This approach could reduce the time regulatory strategists need for impact assessments from days to hours, even minutes. However, GenAI can also produce incorrect and misleading results, also known as hallucinations, that range from simple logic errors to misinformation. Citations included in GenAI results are not granular enough, making it challenging to review accuracy with reasonable effort. As a result, many organizations have decided to wait for foundational LLM maturity before production use in regulatory intelligence automation.

Streamlining document translations

Machine translation tools for consumer use cases have been available for decades. However, recent GenAI innovations have driven the tech sector to heavily invest in other AI forms as well — for example, combining neural networks with traditional machine translation technology for higher confidence level results. This AI-enhanced technology is now entering the mainstream regulatory sector.

Multiple large biopharmas have completed POCs for AI-driven translation to generate submission and label documents for rest-of-world regions using English-based first-wave market documents. These investments are beginning to show proven results in reducing dependence on manual translation service providers and accelerating internal QC cycles.

Leveraging HA query responses

Companies are researching a combination of GenAI and machine learning solutions to generate initial draft responses to HA queries by identifying prior responses to similar questions. This approach has the potential to enable near-simultaneous global product launches and accelerate variation approval timelines across similar product indications and presentations. However, many companies have started to realize that a high-quality HA query database must first be established and maintained to track questions at a granular level before advanced automation can be applied to HA query prediction and response generation.

Generating dossier documents

The most explored regulatory use case of GenAI is automating draft versions of dossier documents. Document types being considered are (a) quality overall summary, clinical summary and non-clinical summary documents in eCTD Module 2 (b) clinical study report (c) safety reports like risk management plan and risk evaluation and mitigation strategy and (d) labels.

A key success factor for this application is sufficient granularity in source document citations and data set traceability for effective review of authored documents. After all, it’s not worth the investment if checking the accuracy of a GenAI-authored document takes longer than manually authoring and reviewing a small update to the previous HA-approved document version. Additionally, not all document and submission types will benefit from this technology because current GenAI models can only summarize existing patterns or insights. Most new drug application submissions and some document types in maintenance submissions will still require a manual review of data sets and source documents to identify new insights and build the story supporting the requested new drug approval.

Evaluating advanced automation investments

Given the technical challenges for viably deploying AI and advanced automation, particularly GenAI at an enterprise scale, regulatory leaders should evaluate and prioritize use cases for all forms of AI (not just GenAI) and non-AI automation investments using a 2×2 matrix of anticipated industry effort versus expected value:

• Effort: Companies tend to focus solely on the explicit effort required in terms of technology costs and business process redesign. However, they should also assess change management, training, and upskilling efforts, including the implicit costs of disruption caused by changes in roles, responsibilities, and the organization’s operating model.
• Value: Companies should assess the impact of advanced automation on workforce productivity, operational cycle times, and strategic business goals, as well as prioritize automation of frequently occurring events across a larger user base. While measuring the long-term compounding impact of continuous investment in advanced technology, also recognize that drawn-out timelines due to phased user adoption can reduce cumulative value.

It’s important to note that not all automation requires AI, and not all AI needs to be GenAI. Basic rule-based automation can provide significant benefits to end-users, especially when adopted consistently throughout the organization. Some use cases include automating document and data quality checks, submission plan refinement, and HA correspondence intake from affiliates and regulatory affairs (RA) strategists.

Setting the tone for disruptive innovation

Sustained investments in GenAI for prioritized, high-impact initiatives will lead to the maturity of foundational LLMs and the development of proven enterprise solutions. For example, LLMs are becoming increasingly secure and could eventually train on anonymized cross-industry data without data leaving company boundaries — an advancement that would benefit the pharma industry at large. Email and Internet search were initially seen as tools for competitive differentiation but eventually became mainstream technologies that shifted entire industries onto different growth trajectories. GenAI has the potential to be the next such innovation.

But before we reach this future state, inaccuracy, non-deterministic results, limited traceability, security, and GxP compliance remain key challenges. As large biopharmas continue to invest in GenAI and evaluate these hurdles, the industry is keeping a keen eye on emerging innovations that could drive the next wave of disruptive regulatory transformation.

Learn about Veeva’s AI partner program, which provides access to critical technology and Vault data to help partners develop GenAI solutions that integrate seamlessly with Veeva applications.