Empowering Clinical Trial Decisions with Data-Driven Decision Management (DDDM)

From chaos to confidence: How i2e and databricks transformed risk based quality management at scale

Risk-Based Quality Management (RBQM), a method of evaluating risks on clinical trials and then applying statistical methods to find outliers at clinical trial sites, has long promised to reshape clinical trial oversight—bringing smarter, faster, and safer decisions to the forefront. But making RBQM operational at scale is hard. Legacy tools, fractured workflows, and inaccessible data often slow progress to a crawl. At i2e Consulting, we specialize in transforming these complex environments into platforms for innovation. For one of the largest global pharmaceutical companies, we helped do just that—modernizing RBQM through a powerful combination of Databricks, Posit, and cross-functional collaboration. The result? Validated, efficient, and collaborative solutions—all delivered under budget. Before: A Fragmented, Manual, and Siloed Ecosystem The client’s RBQM operation was built on legacy infrastructure: Bulk data files were generated by a legacy system and stored on a file server. However, not all needed files were included in the data lake, and others required intensive programming time to prepare. A team of 50 people across the globe, Central Monitors, had to manually open, adjust, and repackage datasets, often in SAS Studio, to get the full view. Supporting analytics relied on desktop Python scripts that only a few could run, due to complex dependencies and inconsistent setup. Data access for additional complex analysis was clunky—long, fragile credential strings and individual workarounds slowed everyone down or made the work impossible to complete. There was no shared development space or common validation pathway, so every new insight felt like a one-off project. The system took years to develop and the client was not able to change it at need. This was working against the goals of a modern RBQM program and the need to be proportional per ICH E6(R3) After: A Validated, Unified Platform Powered by Databricks and POSIT With i2e leading the transformation, the RBQM capability evolved from siloed tools to an enterprise-grade, validated platform: 1. Centralized, Scalable Data Access with Unity CatalogUsing Databricks’ Unity Catalog, our admin granted secure, governed access to the client’s data lake and existing database estate in a single setup. Everyone on the team—from engineers to Central Monitors—could access the same data without redundant effort or manual patchwork. For files missing from the data lake, we uploaded them directly into Databricks, creating a safe and validated environment. We developed a risk based approach to validation with our testing partners. As this is a secondary system, we followed the principles of say what you are going to do, then do what you said you were going to do and document that you did what you said you would do. This is aligned with best practices for risked based validation for clinical trials. 2. Shared, Collaborative Development with Databricks NotebooksWith Databricks notebooks connected to GitHub, development became collaborative, version-controlled, and transparent. The friction of setting up environments disappeared, and code became portable, maintainable, and scalable. From exploratory analyses to formal pipelines, all work could live in a governed, shared ecosystem—one that supported CI/CD (Continuous Integration/Continuous Delivery) for validation-ready software releases. 3. Streamlining Operations with Databricks Workflows Efficient orchestration was critical to transforming the client's RBQM operations from fragmented manual processes into a seamless automated pipeline. Leveraging Databricks Workflows, we built robust job orchestration that systematically generated and refreshed essential data files, eliminating the dependency on manual file preparation. Jobs that previously required manual intervention and extensive coding in SAS Studio or Python scripts were now centrally managed, scheduled, and monitored from a unified interface. Built-in monitoring and alerting capabilities provided immediate visibility into job execution statuses. Any potential issues were proactively flagged, enabling swift troubleshooting and minimal downtime. 4. App Deployment with Posit By connecting to Posit Connect and Workbench, we delivered a suite of workflow apps as part of the“CM Toolkit”. These allowed Central Monitors to interact with complex tools through clean, intuitive interfaces—all without writing code. Additionally, API integration with Databricks Workflows enabled the Posit apps to trigger workflows directly, creating a modular, interconnected architecture that improved responsiveness and flexibility. These apps weren’t in the original scope, but thanks to Databricks’ efficiency and flexibility, and the skills of the i2e team, we delivered them as an add-on feature—while still coming in under 50% of the original budget. 5. End-to-End Integration We leveraged two existing production APIs by creating scheduled Databricks jobs that periodically consumed and integrated the RBQM environment with both the: ticketing system for workflow tracking, and central monitoring platform to contextualize findings. These integrations enabled real-time communication between business and technical systems, improving both insight and responsiveness. 6. Organizational Milestones This became the first validated deployment of Python apps at this sponsor organization. It happened through intentional partnerships—with platform, architecture, and validation teams—and through i2e’s leadership in: Training and enablement Role modeling engineering best practices Building and maintaining complex, production-ready code We gave Central Monitors “training wheels”—letting them safely participate in the tooling ecosystem without needing deep technical knowledge, while still making meaningful progress. Outcomes: Validated Results, Delivered Fast The transformation brought measurable results: 4 validated releases in less than 12 months (Historically, 1 such release a year would have been successful)EDC (electronic data capture systems for clinical trials) onboarding time reduced by two-thirds—from multiple quarters to under 3 months Over 8 workflow applications delivered via Posit Delivered under 50% of budget, with expanded scope More importantly, the organization now has a repeatable, collaborative, and validated RBQM capability—one built for speed and scale. What’s Next: Databricks Apps, GenAI, and Exploratory Acceleration The journey doesn’t stop here. We’re now focused on: Transitioning to Databricks Apps for simplified architecture and streamlined access control—bringing compute, access, and app deployment into a single pane and removing licensing costs. Increasing use of the Databricks Assistant to support Central Monitors and developers alike. Exploratory analysis apps and GenAI use cases that help surface quality signals, generate narratives, and improve efficiency across trials. Most importantly, we hope to showcase a model for others: That business-led and professionally supported software development can coexist—with the right tooling, process, and team culture. And for us, the platform that made this possible is Databricks. At i2e, we bring more than just tools—we bring a blueprint for transforming clinical trial data into a strategic asset. From zip files and scripts to validated apps and APIs, we helped this client move from chaos to confidence. And we’re just getting started.

Change management- What is it? and why should project managers understand and own it

Change management is a critical component for organizations and projects undergoing transformations be it process or technological. Not addressing changes can lead to disengagement, decreased productivity, wasted resources, ultimately leading to project delays. A well-structured change management plan enables organizations to navigate change smoothly, minimizing disruptions and increasing the likelihood of achieving desired outcomes. In this blog we will present a change management guide and explain why it is important for project managers to also become change managers with a realistic example. Before knowing how to master change management, lets know more about it. According to the Harvard Business School, “Change Management refers broadly to the actions a business takes to change or adjust a significant component of its organization. This may include company culture, internal processes, underlying technology or infrastructure, corporate hierarchy, or another critical aspect.” (1) If you are a project manager, you may need to play a dual role of the project manager and a change manager. As change management deals with people’s adoption towards the new process or technology, there is no better person a team trusts than the project manager. So, first let’s understand how different change management and project management are, as this will help you to effectively shift gears whenever necessary. Are Project Management and Change Management the Same Thing? Though there are overlaps between project management and change management, they are two different processes as the former deals with what people are doing and the latter is how the people are doing it – and each of them requires separate approaches. So, how different are change management and project management? Let's find out. By treating change management and project management as separate but complementary disciplines, companies can effectively address both the human and technical aspects of organizational initiatives. Change management is often associated with challenges such as resistance to change, unforeseen risks, data standardization, change fatigue etc., So, how can project managers navigate through these challenges and achieve successful change management? Read the next section as we illustrate change management with the help of a real-time example.Change Management Example: Migrating and Adopting Planisware A global organization recognized the necessity of transitioning its fragmented Project management systems, including Microsoft Project, Smartsheet and Excel, to a unified platform such as Planisware. This is a huge change that the company planned to go with a big-bang approach rather than opting for a pilot project first. Navigating such a change to new technology presents challenges. A robust plan and effective change management by the project managers is required to ensure a smooth transition. Let's begin by looking at the organization’s key objectives the company was looking to achieve with this change. Achieve one tool approach: Different project groups were using both MS Project and Smartsheet. These users loved their tools; however, each project manager utilized them in their own way, with their own tricks and without harmonization in how the data was being collected, structured, and shared. Reports were ad-hoc and as-needed. The MS Project users didn't coordinate with the Smartsheet users, even though some of the same takes and milestones were represented in each. The company aimed to bring all the projects under Planisware and achieve data synergy between teams.Increase data timeliness: Organization wanted to streamline data collection and reporting processes with the help of advanced capabilities of Planisware. They also aimed to reduce the time required to gather and analyze data, ensuring that decision-makers had access to timely and relevant information.Break down silos: Planisware will also foster collaboration and alignment between teams. So, the organization sought to break down silos and promote knowledge sharing. They also wanted to have enhanced visibility into resource allocation and project milestones would facilitate cross-team synergies and alignment with global planning objectives.AI integration: They wanted to get data ready for AI use by centralizing all project data in one repository and standardizing data. Thus, enabling the organization to leverage AI generated advanced analytics for decision-making and forecasting.Resource management: They also wanted to centralize processes for resource management by replacing the offline method. The agenda behind doing this was to enhance visibility, facilitate collaboration and align with global planning objectives. Now let's see how a change management plan can be made for this scenario. Change Management Best Practices for Project Managers In the above example, the company may engage a technology partner to implement Planisware. This partner will take care of the implementation and drive change management. However, there are some crucial aspects where project managers need to switch roles and become change managers as well. Let's explore more of these aspects. Executive push to use data: It is imperative to start including Planisware in the day-to-day activities of the project management. Coach the team members to let go of the previous practices and get data from Planisware only. If there are discrepancies in the data, instruct the team on how to fix them.Practice data driven decision making: Establish data quality control mechanisms to make sure the data is of high quality. Communicate to your team with evidence of how the decision makers are using the data for decision-making, and how the data quality is getting reflected in the reports.Communicate improvements: Gather statistics which can be used to quantitatively show improvement after the change. e.g., number of hours spent on each annual operating cycle, resource utilization, number of projects' gate reviews performed per quarter.Include in the bigger picture: Communicate to the highest-level executive to include updates on the implementation project in all-hands meetings, so that end-users realize that this project is a big deal, and their annoyances are a part of a corporate improvement.Strive to provide material value to your team: Project managers should meticulously plan and work with the Planisware implementation partner to provide material value to each end-user, be it a specific report that will address a challenge or building RPA (Robotic Process Automation) to save them time. For example, building templates for roadmaps for the project managers, so that you can have a one-button operation to output their Gantts into 'pretty' views which can be shared throughout the organization.Meticulous data migration: Rather than lifting-and-shifting all pre-existing data, have the users re-created it in the new tool with the new processes (especially if there are any material changes to the processes, such as new templates, new resource structures, new levels of details expected in their planning). While this seems tedious and users will have some resistance, it is the best form of training and will ensure that they are bought in to the data that is in the new tool (instead of an automated migration process resulting in data that they own but don't trust).Build a data dictionary/glossary: Work with the implementation partner and other stake holders to ensure harmonization or alignment of terms. For instance, in the pharma world, "FPI" sometimes means first subject screened and other teams means first patient first dose. If your organization grew via acquisition, different team members may have different interpretations of the same thing and a data glossary can help establish harmonization.Define resource management terms: Make sure the crucial metrics pertaining to resource management are defined before implementing the module. For example, when the definition of FTE is not clear, then, it’s impact is seen when a report says, "2 FTE are needed for this time period" does it mean that you need 2 humans, or does it mean you need more people because of vacation and trainings and other overhead; can it mean 4 humans working halftime each is ok? etc. And how will finance monetize the estimate for their budget reporting.Identify change agents: Find and enlist change agents, these can be your biggest supporters, but it can also be those most resistant to the change. Maybe it's the admin from the old tool, who is subconsciously fearful for their job. By having them included in the implementation, they can see the value and possibility. Make sure the change agents receive more communication about the change, and be involved in some decision making, so that they are better equipped to support their teams.Setup one functionality at a time: Start implementing Planisware module by module within your team, so you can hold their hand and give them the attention that they need. Similarly, roll out project by project and not a big-bang approach.Conduct “What’s in it for me” training: As a trusted source for your team, the project managers should highlight the benefits of the new tool and process. This requires digging deep into existing ways of working and finding frustrations that people are used to. For example: MS Project has a limited number of baselines. Or double entry was required from a CTMS tool into MS Project (and triple entry then into a reporting tool). Communicate and demonstrate how Planisware can effectively solve this.Prioritize ongoing training: Make sure along with initial training, your team also gets ongoing launch & learn sessions for continued training. Include advanced tips and tricks beyond what is included in the standard training, quick reference guides, like FAQs, are best to support the end users effectively. Introducing a new tool or a process comes with a promise of enhancing the existing processes and eliminating challenges; however, to reap the benefits, change management should be done effectively. Engaging a technology partner for the Planisware implementation makes it smooth and 100% fool proof. At the same time, it is equally important to appoint an experienced team to help with the change management both from the perspective of the tool and the people involved. i2e Consulting has over a decade of experience in PPM and Planisware. We have successfully configured Planisware and aligned the tool to their specific requirements and workflows. Our team also worked closely with the Planisware team and the stakeholders to help with the change management aspect. References 5 critical steps in the change management process; Business Insights; Havard Business School Online

Understanding future of agentic AI in pharma: QnA with Puneet Kacker

“In a world of talkers, be a thinker and a doer.”The difference between generative AI and agentic AI can be rightly explained by this quote. While gen AI is like a thinker helping you ideate and create, agentic AI is both a thinker and a doer. In fact, it goes one step further by initiating and completing tasks and taking decisions autonomously. In the pharma and life sciences industry, this capability plays a key role than one might think of. Imagine using a tool that not only researches but also manages complex workflows. To get more clarity on agentic AI, we spoke with Puneet Kacker, an industry expert and passionate advocate for innovation. He is actively involved in promoting the use of advanced AI technologies in pharma, healthcare, and life sciences, and in driving better and faster solutions to the world. We gained valuable insights into how this next-generation technology is transforming the pharmaceutical industry and how we can work using it to achieve better results. Here’s the blog that shares his thoughts on Agentic AI and its future in pharma and life sciences industry. 1. What is agentic AI and how is it different from generative AI? Generative AI is a model to create novel content, including text, images, audio, video and code, quite efficiently. Primarily, it is an instruction-based AI model that is built on large data sets. Hence, it has a strong capability to understand language prompts and generate high-quality outputs.On the other hand, agentic AI works on a goal-oriented model that operates on high-level instructions, and the agents trigger specific actions. The model uses memory and offers mission-oriented solutions. It can be understood as a human decision-based system that mimics actions typically performed by humans.A quick example: Gen AI can create an email, while Agentic AI can write, send, track the response, and proceed to send follow-ups 2. What are the key challenges pharma companies should overcome to successfully adopt agentic AI into the industry?The pharmaceutical industry has been adopting new technologies at a relatively slower pace. It’s not reluctance, but more due to highly regulated sensitive data involved. Agentic AI holds promise to transform certain workflows; however, there are certain reservations due to the LLM limitations in healthcare. Certain tests results have experienced LLMs to be unsafe and hasty. Hence, one should not rely completely on agentic AI for sensitive decisions. Another challenge is domain knowledge and operations in pharma; specifically in R&D, clinical trials, and regulatory aspects. That said, agentic AI can play a role in specific pipelines related to innovation discovery, but it needs to mature significantly. And even when it does mature, it is still just a tool. If competitors or other companies also adopt similar tools, then they may all end up generating a similar set of results.The key for pharma companies is how to retain human talent as-is, while capitalizing on agentic AI systems in a way that both can be plugged together to bring out the best of innovation. For example, pipelines that collect data routinely, highlight issues in clinical trial data, convert data, transform it, and save it into databases. These routine jobs with complex pipelines can be handled better by using agentic AI. So, in the next 5 to 10 years, these agentic AI systems are expected to become reliable and mature, building a rich storehouse for pharma and similar industries. 3. What measures should be taken for smooth collaboration between human experts and agentic AI agents?Collaboration is critical for the pharmaceutical industry, and human intervention is non-negotiable. Agents work autonomously and it’s difficult (even unethical in some cases) to trust their decisions without cross-checking. Even a minute error in AI-led judgment could lead to a big impact, especially in clinical trials and drug discovery domains where the stakes are too high. Science always believes in understanding the “why” and “how.” Though agents might have logical reasoning behind their outputs, innovation driven by the scientific mindset and human judgement, by core capability, is necessary.Thus, human experts in the pharma industry would require more attention on building awareness about how agentic AI systems work. It’s also essential to ensure that any output produced by the agentic AI system that’s going out of the expected or acceptable range should be flagged.Scientists should understand how the system works and where customization is required. In the current scenario, scientists are crafting prompts and communicating with AI agents, which is a part of shaping innovation. But if AI handles everything, it would be challenging for scientists to make changes that align with their knowledge and intent. To conclude, a lot of learning is required to excel and capitalize on the agentic AI system. 4. What core capabilities make agentic AI suitable for solving the pharma industry’s complexities?Agentic AI can support scientific processes and due diligence, especially in forming hypotheses and bringing relevant data to scientists in less time. A lot of curations, which have traditionally been done manually, can now be handled faster using AI. However, it comes with a lot of reservations. If mistakes in the system go unidentified and are released, they could have detrimental effects.So, human oversight, with domain knowledge and technological expertise, will continue to play a critical role. Routine tasks such as data collection, data cleaning pipelines, and preparing data for analytics can be handled effectively by agentic AI systems. That said, there are some reservations about its use in clinical trials, especially in handling patient data, monitoring disease progression, and clinical trial recruitment. So, strong human oversight will always be necessary. 5. What key factors should pharma companies consider when implementing agentic AI? Figure - Key factors agentic ai 6. Will agentic AI find greater adoption in project management domains within the pharmaceutical industry?Agentic AI does have a lot of potential since the data is structured, and the rules are well defined. It has the capacity to consolidate data from multiple sources and coordinate complicated schedules. That said, agentic AI cannot likely replace project coordinators. It works best when collaborating with human decision makers who can bring in the right judgment. Organizations will require people who are domain experts and also have a strong understanding of advanced technologies like agentic AI. These individuals can customize solutions, flag issues, and take control when needed. It’s just like a self-driving car; there’s always a manual override option. Likewise, in project management, human oversight will always be essential. Leaving critical operations solely to agents could risk project failure or even damage a company’s reputation. Conclusion While agentic AI holds immense scope for speeding up processes in the pharmaceutical industry, it’s clear that human oversight along with deep human intelligence will remain the key elements. Dr. Kacker strongly believes, “the future is AI with human, but not AI vs human.”Pharma faces constant innovation challenges, with new diseases and evolving problems that require adaptable, domain-specific expertise. AI might solve specific problems efficiently, but human capability is what brings flexibility, context, and foresight. The future of pharma, therefore, lies in a strong, complementary partnership between human expertise and machine intelligence, working together to drive meaningful progress .—a progress filled with hope and support for patients around the world.