Accelerating time to market: leveraging AI to build process intelligence and streamline drug development

By Unjulie Bhanot, Product Marketing Manager, Process Development & Manufacturing, IDBS

Blog 3 of 3. Read blog 1 and blog 2 here.

BioPharma organizations face long clinical development times, with the median development time for innovative drugs standing at 8.3 years. Additionally, 9 out of 10 drugs fail clinical trials. To expedite development, fail faster and deliver life-changing therapies to patients more quickly, a robust pipeline of targeted molecules and effective tools are essential. AI offers significant promise in this race. The current global market size for AI in drug development is over $1.5 billion, with a forecasted value of $7.8 billion by 2032, growing at a compound annual growth rate (CAGR) of 19.4%. AI can accelerate drug discovery, improve drug design, enhance clinical trials and ensure safety monitoring, ultimately reducing time to market.

At the heart of AI’s ability to deliver reliable results for the BioPharma industry is high-quality data—data that is accurate, contextualized, consistent and complete. Data that enhance product and process understanding to build process intelligence. To build these comprehensive, trustworthy datasets, BioPharma organizations need to deploy a data infrastructure that consolidates all the relevant data, to drive data-informed decision-making.

Within process and clinical development, as discussed in our earlier blogs; to achieve these gold-standard data sets, it is critical to improve digital maturity and leverage the right digital tools to build product and process knowledge. This lays the groundwork needed to unlock the potential of AI tools.

The potential of AI in BioPharma

AI encompasses the use of data and algorithms to enable computers to perform cognitive functions we associate with human minds, such as learning, reasoning, problem-solving and decision-making. Machine learning is a major subset of AI capabilities, within which reside deep learning capabilities. Generative AI (GenAI), a form of deep learning, is expected to transform the BioPharma industry, potentially generating $60-110 billion in annual value across the value chain, with $17-32 billion of this value within clinical development and operations.

To reap the full value of AI, comprehensive digital transformation across the value chain is essential. Without a clear strategy, and the end of siloed, decentralized software deployments, its value is limited to the short term, and often quite lackluster. It cannot be clearer, that to leverage and scale AI, organizations need to adopt a data-driven platform-powered approach across the ecosystem, to unify data, maintain data integrity and create reliable comprehensive data sets.

Large language models (LLMs), a subset of GenAI, hold significant promise in BioPharma. Designed to process and generate text like humans, LLMs can interpret and respond to questions, analyze data sets and discern patterns and trends. They are proving particularly useful in the drug discovery space – extracting scientific knowledge from scientific papers to contributing to the design of drug molecules by generating molecular structures with desired properties. But how can we look to leverage these models in the development space, to streamline product development, smarten process monitoring and build process intelligence?

Process intelligence can serve development and manufacturing

Process intelligence leverages data and advanced analytics to help process development and process characterization teams gain a deeper understanding of their processes. This enables them to determine how to control sources of variability, improve efficiencies and ultimately optimize processes to deliver high-quality product, facilitating smoother technology transfer and de-risking manufacturing operations.

During early and late-stage process development, teams are building product and process knowledge by integrating experimental, process and product data, often residing in disparate, legacy systems (or even in the mind of a scientist). Critical errors, loss of context, or data omission can occur during data assimilation. Implementing AI tools under these circumstances can yield poorly trained models and unreliable analysis – certainly hindering process intelligence and predictive capabilities.

However, the value of building this process knowledge extends beyond process development – if process knowledge is sufficiently built during process development, the process intelligence it yields can serve the manufacturing team. Maintaining a detailed record of process parameters and their evolution in development can make it easier to replicate and transfer processes during tech transfer and provide critical insight during faults and failures in manufacturing runs. True process intelligence provides an understanding of all components of a drug development process – insights into the molecule itself, the impact of all contributing materials and recipes, and the decisions behind what determined the criticality of a given process parameter. In turn, this helps manufacturing teams improve process predictability, reduce run times and deliver high-quality, high-yielding product.

Enabling process intelligence with complete, contextualized data

To leverage GenAI for process intelligence, the underlying model must be sufficiently trained. For example, being able to ask the AI agent the ideal upstream and downstream development process for a novel IgG1 for a breakthrough cancer treatment, to create sufficient material for Phase 1 trials, relies on high-quality data from a previously developed, well-documented process for IgG1 manufacturing.  The more detailed the data, the better the AI agent’s response.

Training an LLM model for process development is perhaps similar to preparing a lasagna – just as you’d prepare a ragù in one pan, a béchamel in another, and the lasagna sheets in a separate pan and piece them all together in the roasting pan to go in the oven –  to effectively train an LLM model for process development requires organizations to access current and historical execution data (process development experimental records), process and parameter design data (key and critical process parameters, and critical quality attributes), and product data (result value from analytical assays and stability studies).

In today’s ecosystem, execution data may reside in an ELN, and product data may be in a LIMS system, but process data may be on paper, Excel or an orphaned document scribbled on by a scientist. The ELN and LIMS will store data in their own database, and the process data? That will require manual transcription into an electronic system to digitize it, storing it in a third database. Consolidating this data provides the necessary context for accurate product and process intelligence, though it is a manual, risk-prone process.

The rise of the one-pot dish

Traditionally, lasagnes were prepared in separate pots and then combined to create the final dish. However, we have now entered the era of “one-pot” dishes, offering benefits from convenience to flavor integration, efficiency and cost-savings.  This approach can also benefit BioPharma – swapping convenience for streamlined process execution, flavor integration for enriched data context, and also yielding process efficiencies and development cost-savings.

To build process intelligence, BioPharma stands to gain from the one-pot approach – bringing experiment, product and process data together to concurrently build product and process knowledge, streamline process development and execution, and unify data with context to create complete, contextualized AI-ready data sets. This enables teams to move from process knowledge to process intelligence, optimizing their processes for the best performance. It is time for teams to determine how their processes run today, optimally design future processes and ultimately monitor, control and optimize their processes for the best process performance.

One software to bring processes, experiments and products data together

Over this 3-part blog series, we discussed the roles of experiment-focused and sample-focused informatics solutions in the drug development lifecycle and uncovered the gap in process-focused digital tools that can manage parameters and processes and determine the evolution of parameters and processes over time. To address this, IDBS has released Polar Control, a groundbreaking module within the Polar platform. Polar Control helps teams manage and evolve their processes in situ with experimental execution and product management, bringing together data with context, driving process excellence with parameter control and traceable experimental results, and ultimately creating the comprehensive datasets AI tools need to accelerate data-driven insights and build process intelligence. One software platform—one pot—to help accelerate time to market.

Learn more about Polar and Polar Control and see how IDBS can help you deliver life-changing therapies to patients faster.

Read blog 1 and blog 2 here.

 

About the author

By Unjulie Bhanot, Product Marketing Manager, Process Development & Manufacturing, IDBS

Unjulie Bhanot is the Product Marketing Manager for Process Development and Manufacturing at IDBS. With over 10 years of experience in the Biopharma informatics space, she has led the strategy and development of IDBS’ Bioprocess solutions and was instrumental in the launch of IDBS Polar to market.

 

 

References: 

1. National Center for Biotechnology Information. (2023). Clinical development times for innovative drugs. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC9869766/#SD1
2. National Center for Biotechnology Information. (2022). Why 90% of clinical drug development fails and how to improve it? Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC9293739/#:~:text=However%2C%20nine%20out%20of%20ten,and%20drug%20approval2%2C3
3. DataIntelo. (2025). Global artificial intelligence drug development market. Retrieved from https://dataintelo.com/report/global-artificial-intelligence-drug-development-market
4. McKinsey & Company. (2025). Generative AI in the pharmaceutical industry: Moving from hype to reality. Retrieved from https://www.mckinsey.com/industries/life-sciences/our-insights/generative-ai-in-the-pharmaceutical-industry-moving-from-hype-to-reality
5. IDBS. (2025). Unlock efficiency: Leverage the right tool for process knowledge. Retrieved from https://www.idbs.com/2025/01/unlock-efficiency-leverage-right-tool-process-knowledge/
6. IDBS. (2025). Process-focused digital solutions: Redefining process design and the evolution of critical process parameters. Retrieved from https://www.idbs.com/2025/02/process-focused-digital-solutions-redefining-process-design-and-the-evolution-of-critical-process-parameters/
7. IDBS. (2025). IDBS Polar. Retrieved from https://www.idbs.com/polar/
8. IDBS. (2025). IDBS Polar Control: Drug development’s recipe book for efficient parameter and process lifecycle management. Retrieved from https://www.idbs.com/polar/process-knowledge-for-better-process-control/