Agentifying post-approval change management: From change to defensible regulatory strategy
8 min de lectura

At Freyr One USA 2026, the session ‘Agentifying post-approval change management’, presented by Wajeed (Senior VP, Freyr Solutions), explored how agentic AI can help regulatory teams manage the complexity of maintaining approved products across global markets. 

The session began with a simple point: approval is not the finish line. For global pharmaceutical companies, keeping an approved portfolio compliant, supplied and commercially viable requires continuous regulatory vigilance. Every manufacturing update, quality change or new regulatory requirement can trigger a chain of assessments involving multiple products, registrations and markets. 

Through the session, the role of specialized AI agents in connecting product data, health authority requirements, dossier information and historical precedents came into focus. These capabilities can support more efficient, consistent and defensible regulatory decisions while keeping human experts in control. 

Why post-approval change management consumes so much regulatory capacity 

For large pharmaceutical organizations, post-approval activities can consume as much as 80% of regulatory resources, according to estimates presented during the session. A large global pharma company may manage close to 300 post-approval changes annually, potentially translating into as many as 6,000 market-level change assessments. 

Each change raises questions about affected products, registrations, markets, filing requirements and supporting documentation. Addressing these may involve more than ten cross-functional teams, with some changes taking around 18 months from initiation through implementation. The challenge is not simply managing submissions, but making consistent regulatory decisions across a complex portfolio while avoiding missed impacts, delayed filings and potential disruptions to supply. 

Connecting changes with the right products, registrations and markets 

A change raised in a quality system may appear routine, but its regulatory implications are not always immediately clear. A new lot of a reference standard, for instance, may affect several products using the same testing methodology, even when those products are not explicitly identified in the original change record. 

This is where connected regulatory information becomes important. Freya brings change information into the regulatory change management process and uses product data, registration records and related information to identify potentially affected products and markets. It can recommend additional products based on shared characteristics, flag them for expert verification and distinguish active registrations from those that have been withdrawn. 

These capabilities help establish the scope of an assessment before regulatory teams begin evaluating the filing requirements. Rather than relying entirely on manual investigation across separate systems, teams can work from a consolidated assessment pack, with relevant products and registrations brought together for review. 

Three assessments provide the basis for regulatory decisions 

Once the affected products and registrations are identified, the next step is understanding what the proposed change means for each market. The session highlighted three complementary assessments that help establish a more complete regulatory picture. 

Dossier assessment examines what was previously filed and approved, identifying whether the proposed change affects information within the registered dossier. 

External assessment evaluates applicable health authority requirements using Freyr's Granular Regulatory Requirements (GRR) database to determine the relevant filing obligations. 

Precedent assessment reviews historical submissions and previous regulatory decisions involving similar changes to provide additional context. 

Together, these assessments help regulatory teams determine the appropriate course of action for each registration. The same change may require different filings across markets, depending on what was originally approved, the applicable health authority requirements and previous regulatory interactions. Bringing these sources together allows recommendations to be supported by relevant evidence, rather than relying on isolated information or historical practices alone. 

Human judgment remains central to agentic change management 

One of the key messages from the session was that AI agents support regulatory decisions, but do not replace the experts responsible for making them. 

The same post-approval change can lead to different regulatory outcomes across markets. While the dossier information, health authority requirements and historical precedents may support a consistent filing recommendation for one registration, another may require further clarification or expert review. 

In the case explored during the session, the three assessments supported a consistent variation recommendation for an EU registration. However, for a US registration affected by the same change, the assessment required confirmation of whether a qualification protocol had been approved. Rather than making an assumption, Freya sought clarification from the relevant product lead. The response helped establish the applicable regulatory requirement and filing recommendation. 

Historical precedents can introduce further complexity. A similar change filed previously may suggest one approach, but its regulatory outcome may tell a different story. In this case, an earlier submission had been filed as an annual report but subsequently received a follow-up question from the FDA, requiring updates to the dossier. This additional context helped the product lead evaluate the precedent and support a different filing recommendation. 

The expert's assessment, supporting evidence and reasons for the decision were then brought together into a consolidated report. Freya helped compile the information, while the product lead reviewed and approved the final recommendation. 

This human-in-the-loop approach is particularly important when dealing with exceptions, conflicting precedents or incomplete data. AI agents can perform much of the assessment work and propose a course of action, while regulatory professionals provide the interpretation and final judgment. 

The result is not simply a faster assessment, but a decision supported by documented reasoning that can be reviewed and traced. 

Specialized agents connect impact assessment with filing strategy 

The session also explained how specialized agents work together within a broader six-step post-approval change management process. 

Freyr's approach covers the journey from identifying a change to developing the regulatory filing and content strategy. The six steps are: 

  1. Change initiation and assessment: Capture the proposed change and establish its initial context. 
  2. Impact assessment: Identify the affected products, markets and registrations. 
  3. Variation identification for individual registrations: Assess the regulatory implications and applicable variation requirements for each registration. 
  4. Variation identification across multiple registrations: Consolidate and evaluate variation requirements across affected markets. 
  5. Submission strategy and grouping: Determine filing priorities, timelines and opportunities to group related submissions. 
  6. Content and submission alignment: Identify the dossier content and documentation that must be updated or prepared for submission. 

Specialized agents support different stages of this process. The Product Impact Assessment Agent (PIA) helps identify products potentially affected by a proposed change, while the Market Impact Assessment Agent (MIA) determines where those products have active registrations and which markets may require regulatory action. 

The Variation Impact Assessment Agent (VIA) builds on this context by using structured regulatory requirements to assess the nature of the change, support variation classification and identify potential filing obligations across markets. 

Together, these agents connect product and market information with regulatory requirements to support a more coordinated assessment. The objective is to move beyond identifying what has changed and establish what action is needed for each affected registration. 

Connecting existing systems and regulatory information 

The audience raised a practical question: how can agentic change management work with their existing systems? Change-control information often comes from different vendors, manufacturing partners and internal systems.  Formats may vary, descriptions may be inconsistent and individual changes may be linked to earlier records. 

The approach discussed was to bring this information into a connected data layer, allowing AI to interpret change descriptions alongside historical assessments and regulatory records. This does not necessarily require replacing existing systems, although reliable integrations, data traceability and compliance remain important considerations. 

Classifying existing dossier documents and extracting relevant metadata can further strengthen this foundation. Keeping that information updated as new submissions and approvals occur supports more reliable regulatory assessments. 

The next stage of agentic post-approval lifecycle management 

While the first four stages of the six-step process focus on identifying and assessing regulatory impacts, the final two extend into filing strategy and content strategy. 

Filing strategy involves prioritizing submissions across markets, considering regulatory timelines and identifying opportunities to group related variations. Content strategy goes further, examining what needs to be modified within the dossier and which information must support the proposed change. 

These capabilities were discussed as areas of ongoing development, with further filing-strategy and content-strategy functionality targeted for December 2026 and March 2027, respectively. 

The longer-term opportunity lies in connecting these stages rather than automating them independently. For organizations managing large global portfolios, this could reduce repetitive assessment work, improve coordination across functions and give regulatory professionals more time to focus on decisions that require their expertise. 

Looking ahead: a more connected approach to post-approval change management 

Post-approval change management involves far more than identifying a variation and preparing a submission. It requires continuous coordination between regulatory requirements, product information, approved dossiers, historical decisions and the teams responsible for taking action. 

The session highlighted how specialized AI agents can support this process by bringing relevant information together, assessing potential impacts and preparing recommendations for expert review. 

As these capabilities evolve, their value will depend on how effectively they connect with existing systems, regulatory processes and human expertise. The objective is a more coordinated approach in which agents handle much of the assessment effort while regulatory professionals remain responsible for the decisions that matter. 

Explore how Freya Fusion brings regulatory data, specialized AI agents and connected workflows together to support post-approval change management. 

Discover Freya Fusion 

Preguntas frecuentes

AI is used to track and assess regulatory changes, compare and classify regulatory data and documents, support post-approval change assessments, and assist with drafting. Its value grows when it is built into enterprise processes rather than used as a standalone tool.

A single post-approval change can affect multiple products, registrations and markets, each with different regulatory requirements and filing obligations. Managing these dependencies requires coordination across regulatory, quality, manufacturing and other functions. As portfolios expand, maintaining consistent assessments and timely submissions becomes increasingly challenging.

A defensible assessment considers what was previously approved in the dossier, the applicable health authority requirements and relevant historical regulatory decisions. Bringing these sources together helps establish the rationale behind a filing recommendation. Documented evidence, traceable reasoning and appropriate expert review are essential when assessments involve incomplete information or conflicting precedents.

Agentic AI can connect product and registration information with regulatory requirements and historical records to support assessments across multiple markets. By applying a structured approach to identifying impacts and evaluating filing obligations, it can reduce repetitive manual work while accounting for market-specific differences. The objective is consistent assessment, not necessarily identical regulatory outcomes.

No. AI agents help gather evidence, perform assessments and prepare recommendations. Regulatory experts remain responsible for validating information, resolving exceptions and approving final decisions, particularly when requirements or historical precedents need further interpretation.

AI-enabled change management can draw on information from existing quality, change-control, regulatory and document management systems through appropriate integrations. Connecting these sources allows organizations to use their existing regulatory information without necessarily replacing established systems. Data quality, interoperability, traceability and compliance requirements remain important considerations.

Beyond identifying regulatory impacts and variation requirements, agentic capabilities are being developed to support submission prioritization, grouping and dossier content planning. Connecting these activities could help organizations move from individual assessments toward a more coordinated approach to post-approval lifecycle management, with greater visibility across regulatory processes.

Suscribirse al blog de Freyr

Política de Privacidad